Skip to content

feat(dashboard): Dashboard edit mode, see where a value came from and ask the Bot to change it - #33265

Closed
chenmingwei23 wants to merge 2 commits into
mainfrom
feat/dashboard-edit-mode-lineage
Closed

chenmingwei23 wants to merge 2 commits into
mainfrom
feat/dashboard-edit-mode-lineage

Conversation

@chenmingwei23

@chenmingwei23 chenmingwei23 commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: On a Bot's Dashboard tab, the person can see where any value came from, click a component to ask the Bot to change it, and correct a value the Bot wrote, all behind an opt-in feature preview.

A Dashboard page shows numbers with no way to ask "where did this come from?" from the page itself. To change one card the person has to describe it in chat from memory. A value the Bot wrote cannot be fixed by hand. Pages that draw in script (canvas, WebGL, script-written text) and pages fed by a script cron (a CloudWatch metric, a PR count) are the hardest case: nothing the eye reads is a bound element.

Why it matters

The person trusts a number only when they can trace it, and asks for a change faster when they can point at it. Without lineage, a stale or wrong value looks the same as a current one, and a script-only board gives no hint of which job, script or reading produced it.

Not a goal

  • A stable board key across reads, crew-log seq on fold rows, evidence on verdict entries, and source / observed_at on dashboard_write. Each shows in the lineage as a needs: <what> hop instead.
  • A declared derive on a model field (a page's own math).
  • Lineage for a root session's own page; this ships for a Bot's page only.
  • Changing the page shape: components are named regions inside one page; no card grid is forced.
  • Locking a corrected value: the Bot may write over a person's correction, and its next turn is told the person corrected it.
  • Letting a page script trigger Edit, a change request or a correction.

What changed (motivation → approach → change)

Approach: the host draws every word of a lineage, and the page only says which field path is under the pointer, so a page can point at a field but cannot fake where it came from.

  • Preview dashboard_edit_mode (default off): a Settings card, a gateway flag, and a browser flag. Off, the composed document, every package write and fold row, and the Bot's turn are byte-for-byte as before (pinned by tests).
  • Component ids: view.blocks[].id becomes data-dashboard-block; ids are checked (charset, unique, at most 64) at the package write gate only, so a stored page never stops reading.
  • Frame bridge: a script composed before the bootstrap answers host point messages with {path, rect, block} for bound cells, and through kirocrew.pick(handler) for script-drawn values; keyboard focus reports too. Its capture listener stops edit / point / wheel before the page sees them.
  • Lineage route GET /api/members/{slug}/dashboard/lineage: owner-only, read-only, path validated; hops from the field spec, the fold value, the agentic row (writer, time, earlier write, and freshness from the field's stale_after_s through dashboard_agentic.stale_after_map, the same time the page's out-of-date mark uses) or the custom record (writer, stale_after, last failure, gateway stamp, script's declared source).
  • Host UI: an Inspect button (the mode inspects first; the preview keeps its name, Dashboard edit mode), capture layer, lineage popover (connector line, hop-by-hop, reduced motion instant, Esc, never covers the value), a "Change {component}" card that turns over from the clicked component, with provider and a change request sent as the person's own chat message with a context line, Sources legend.
  • Popover Edit (pencil or E; the popover pins while Edit is open; the first Esc closes Edit, the second the popover). Every value: ask the Bot to change it. A value the Bot writes: correct it through POST .../dashboard/correct (owner-only, refused to an agent caller, same type check as dashboard_write, scalars only), recorded with actor: person, the replaced write kept as prior, and a [DASHBOARD] line on the Bot's next turn. A script's record: the job behind it (schedule, hold time, script, last run), Open Schedule, and Run now, which asks first. A folded value: ask the Bot to change how it is shown. Each option that does not apply is shown greyed with its reason.
  • Script provenance: ctx.crew_log_append(..., source={kind, ...}, points=[[t, v], ...]) with kind one of cloudwatch / http / sql / prometheus / file / other, at most 2 KiB, console_url https only and never fetched (a link the redactor rewrites is dropped, not the record); at most 120 points. The gateway stamps the job, run time and script path. It records no digest of the code, because the file on disk when the record arrives may not be the code that ran. The scheduler names no run id yet, so the lineage shows a needs: run id and code digest hop instead. Each hop is marked "recorded by gateway" or "declared by script".

Open questions (each took the conservative option; reversible)

  1. Does a person's correction lock the field against the Bot? No lock: the Bot's next write wins, the lineage keeps both writes, and the Bot is told the person corrected it.
  2. Must the Bot reply when told of a correction? No: it is told, and asked to say why if it writes over the value; nothing waits on a reply.
  3. Should Run now on a job ask first? Yes: it asks, because a run costs one (and an API call).
  4. Should a formula change apply at once? No: the page owns its math; any change goes to the Bot as a request. (Declared formulas are not in this PR.)
  5. Can a person correct a script's value by hand? No: it is measured; Edit offers the job behind it instead.

Other decisions taken while the owner was away: the page stays one page with named regions; only the person can Edit, ask or correct, never a page script.

Backwards compatibility

Compatible: every new behaviour but the row charges below is behind the preview, which is off by default; new crew-log fields are optional and absent with the preview off.

Removes nothing: the change adds a preview, two routes and optional fields; no path, key or validator that works today changes.

One change applies with the preview off: the warm-fold memory charge per row rises for the custom fold (8,900 to 18,100 bytes) and the agentic fold (8,400 to 18,700 bytes), so the charge covers the widest row those folds can keep. Under a fixed slot_fold_cache_bytes ceiling, fewer of those cells are kept warm, and one that is not warm is folded again on its next read.

Tests

  • test/test_dashboard_edit_mode.py (46): preview off keeps the document, blocks and fold rows unchanged; block ids on/refused/duplicate at the write gate only; bridge order and the pick anchor; field-path grammar; lineage for fold, agentic (person write over a Bot write) and script records; the folds keep actor/prior and source/points/provenance; the record door's checks and the gateway stamp; both routes 404 with the preview off; the Bot's turn line.
  • website/src/pages/members/dashboardEdit.test.ts (6): manifest shapes, picked-message sanitizing, refusing undeclared paths, popover placement, correction parsing, change-request text.
  • website/src/pages/members/CrewDashboardEditLayer.test.tsx (6): Edit is posted to the frame, an undeclared path is refused with no gateway call, the gateway's lineage is drawn and a folded value's Edit greys Correct, Edit pins the popover and Esc closes Edit before the popover, Run now asks before it runs, and a change request is sent only on Send.
  • Ran the directly affected suites serially: test_dashboard_frame.py, test_artifact_dashboard_package.py, test_feature_previews_config.py, test_config_baseline.py, test_crew_log_types.py, test_agent_panel_routes_record.py, the configuration doc tests, and vitest catalogParity, previewSurfaces, previewFlagServerSync, featurePreviewsSectionNotice, CrewDynamicDashboard.

Proof: not provable -- other: feature PR, every new test fails on main because the code it calls is new

Manual verification

Screenshots below come from an isolated pod of this branch with the preview on.

Screenshots / video

These frames come from the built SPA (website/dist at this head), driven by headless Playwright. No server or socket was used. Every request to a fake origin is answered by page.route: static files come from the build, and the gateway's /api routes are stubbed. The page payload and every lineage answer are generated by the gateway's own code (dashboard_frame.compose with the edit bridge, and member_dashboard_lineage.build_lineage over seeded fold rows), so their shapes are the real ones. The Bot is "Atlas" with the preview on. Light and dark are shown for every step.

The stubs cannot show steps that need a real Bot turn: the change request arriving in the Bot's chat and the Bot answering it, a corrected value being written and redrawn, and a script job actually running. Those are left for the narrated end-to-end video recorded in a pod.

The change card, recorded: point at p99, click its component, and a card titled "Change latency" turns over from that box. Closing it shows the page again.

Change card recording

The Inspect button is shown at rest, with no hover needed:

Inspect button at rest, light
Inspect button at rest, dark

Inspect mode on:

Inspect mode bar, light
Inspect mode bar, dark

Point at a counted value: one plain sentence, with the steps closed under Details:

Counted value, light
Counted value, dark

Details open on the same value:

Counted value details, light
Counted value details, dark

A script-written value (CloudWatch p99), with its record, freshness, script run, what the script read and its readings:

A value past its stale limit says so under the summary and on its Freshness step ("Out of date: went stale 23h ago."), never as a raw timestamp:

Script record lineage, light
Script record lineage, dark

The Edit button is named for what its menu offers on each value: "Correct it or ask Atlas..." on a value the Bot writes, "Ask Atlas or see the job..." on a script value, "Ask Atlas to change this..." on a counted one. On a script value, correcting it by hand is unavailable, with the reason shown, and the job behind it is offered:

Change menu on a script value, light
Change menu on a script value, dark

The job panel (job, schedule, how long the value counts as current, shown as "2h", script, last run, Open Schedule, Run now). The actions stay pinned at the popover's bottom while the steps above scroll:

Job panel, light
Job panel, dark

Run now asks first ("Run the job once now, outside its schedule?"):

Run now confirm, light
Run now confirm, dark

After the confirm, the run is started (the run route is stubbed), and Run now stays off so it is not started twice:

Run started, light
Run started, dark

A value the Bot wrote, showing who wrote it, the person's earlier correction with its reason, and freshness:

Agentic value lineage, light
Agentic value lineage, dark

Correcting that value. The new value, the optional reason and "Save correction" stay fully in view at the bottom:

Correct a value, light
Correct a value, dark

Click a component: the card titled "Change latency", which shows the provider and the context chips:

Change card, light
Change card, dark

A request typed, with Send enabled:

Change card with a request, light
Change card with a request, dark

After Send, the request shows in Atlas's open chat at once as the person's own message, and the card stays on "Sent to Atlas." until the person presses Done:

Sent, light
Sent, dark

Sources: every field the page reads, with how each is fed:

Sources legend, light
Sources legend, dark

The preview's card in Settings, under Developer and then Feature Previews:

Settings preview card, light
Settings preview card, dark

Related Issues

Part of #33255

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 8, 2026
@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 81c48b84a25cae130bc794595cf4fe8b204c7b61 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: the preview is off by default and gates every new surface, the evidence is attached, and the one change that applies with the preview off (larger fold row charges) is reversible.

Notes

  • "Removes nothing" is not quite accurate. With the preview on, canonical_package_content refuses a view.blocks[].id it used to accept, so a Bot's existing writer that sets uppercase or odd ids starts getting 400s. That is reversible by turning the preview off, but it is a take-away, and the section should list it as one.
  • The custom and agentic fold row charges roughly double for every user, preview on or off, to make room for fields that only exist with the preview on. A conditional charge would keep the cache footprint unchanged when the preview is off.

[DESIGN-REVIEWED] 81c48b8

Description read

[DESCRIPTION-READ] 9fa8ba965650889733c1505ed38d10219ee3fe43fd0750f38fefd7fdd31146fd

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)

UX-level review of 81c48b84a25cae130bc794595cf4fe8b204c7b61 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: PASS

cannot evaluate: missing screenshots of the popover's current Edit button and correction form. The attached images come from an earlier revision, and the only recording appeared blank to the reader.

Evidence gaps

  • Popover Edit button. The code now labels it "Change this value…" with an E key hint (CrewDashboardEditLayer.tsx:1277). Every screenshot shows the older labels instead: "Ask Atlas to change this…", "Ask Atlas or see the job…" and "Correct it or ask Atlas…". The blind read never mentions the current label. This button is how a user reaches ask, correct and see-the-job. To close: new shot-06/10/20 equivalents taken at 81c48b8.
  • Correction form. The form now shows "Replaces the value on the page. {{bot}} is told on its next turn." No screenshot shows it. In shot-22 the reader asked "whether Atlas is told about it and whether I can undo it", which this line is meant to answer. The form's page_changed_confirm / paused_newer_page states appear in no screenshot either. To close: a fresh shot-22 plus one shot of each blocked state.
  • Change card after sending. The code hides Cancel once a request is sent ({!delivered && …}). Shot-28 still shows Cancel, and the reader "would not press Cancel, in case it un-sends". To close: shot-28 retaken at HEAD.
  • Inspect strip text. Shots 04, 06 and 12 show the old wording "Click a component to ask Atlas to change it". HEAD only has "…write {{bot}} a change request. Nothing is sent until you send it." The reader flagged the wording flipping back and forth. To close: a single set of shots taken at HEAD.
  • Change-card transition. The card is meant to turn over from the clicked component. The only recording (shot-01.gif) looked "Completely white" to the reader, so nothing shows the transition. To close: a recording that plays, of Inspect → popover → Change card.

Notes

  • Shot-10 shows the source as "script" while the page label says "(CloudWatch)", and the reader asked which is the real source. Show the script's declared source.kind on the summary line.
  • Mono hop text like "custom → latest.p99.data.ms" and "type number" read as noise to the reader. Show the friendly label first and put the raw path inside Details.
  • Shot-28's "Last turn was interrupted — press Resume" right after a send made the reader think their request failed. Check that a successful change request doesn't leave this banner showing.

[UX-REVIEWED] 81c48b8

Description read

[DESCRIPTION-READ] 62dba26d91a7b5e0263b123b03edfb936aff2ed70ddc20437b1c6e45b0d5a4d2

That sha256 names everything this verdict read: the PR description, plus the 36 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 81c48b84a25cae130bc794595cf4fe8b204c7b61 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

Everything new sits behind a preview that is off by default. The one change that applies with it off, the larger fold row charge, keeps the existing upper-bound rule for rows read off disk.

Notes

  • Item 7 — oversized: the gateway keeps up to CUSTOM_POINTS_LIMIT = 120 [t, v] readings. The only reader, Spark in CrewDashboardEditLayer.tsx:842,917, throws away t and draws points.slice(-40). So 80 readings and every timestamp are stored and charged for in the custom fold's 18,100-byte row, and nothing displays them. Subtraction: cut CUSTOM_POINTS_LIMIT to 40 and lower the custom charge in projection_budget.py to match.

What this change ships

Inventory (9 items) — 8 justified

Intent: Let a person trace a Bot's dashboard value back to its source, ask the Bot to change a component, and correct a value the Bot wrote. ADDITION.

  1. New Settings preview "Dashboard edit mode", off by default — justified
  2. An Inspect button and capture layer on a Bot's Dashboard tab — justified
  3. A lineage popover drawn by the host from GET .../dashboard/lineage — justified
  4. A "Change {component}" card that sends a change request as the person's own chat message — justified
  5. A person can correct a value the Bot wrote (POST .../dashboard/correct), and the Bot's next turn is told — justified
  6. Pages can mark components with data-dashboard-block ids, checked at the write gate — justified
  7. Script crons can declare source= / points= on ctx.crew_log_append — oversized (120 readings stored, 40 values shown)
  8. The gateway stamps each script record with its job, run time and script — justified
  9. Larger memory charge per row for the custom and agentic folds, even with the preview off — justified

[FIRST-PRINCIPLES-REVIEWED] 81c48b8

Description read

[DESCRIPTION-READ] a9d8892d6aa32578e9f0e8fedf22c2c4b27e2ac616fbf59fd59a03e95d8ce776

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — 🔴 changes requested (blocking)

GPT 6.1 found at least one blocking issue that must be resolved before merging 81c48b84a25cae130bc794595cf4fe8b204c7b61. Adjudication upheld 1 of 1 blocking finding(s); those must be resolved before merging.

This comment is updated in place on each push.

BLOCKING -- website/src/pages/members/CrewDashboardEditLayer.tsx:638 -- Retained corrections remain enabled after a field becomes non-scalar

: edit.kind === 'correct' ? undefined
Trigger: With a scalar field’s correction open, the Bot applies a page changing that agentic field to array. The refreshed lineage reaches EditPanel with scalar: false; “Use this page” adopts the array type, and entering [] enables Save.
Outcome: parseCorrection accepts the array and calls memberDashboardCorrect, but the correction route rejects it with 400 not_scalar. The user gets an enabled correction action that cannot succeed instead of the non-scalar explanation provided when opening Edit normally.
Anchor: defect
Fix: Block retained corrections when edit.scalar is false, using no_correct_shape while preserving the draft.
Sites: none

[BLOCK-MERGE] 81c48b8
[GPT-REVIEWED] 81c48b8

Adjudication (Opus 5.5) — is blocking on each finding proportionate?

F1 — LOW harm: the gateway refuses the save (src/kiro_crew/dashboard/handlers/member_dashboard_lineage.py:578, not_scalar), and the box shows a failed-request notice (CrewDashboardEditLayer.tsx:1061-1066). Nothing is written or corrupted, and the draft is kept. The condition is real: blocked at CrewDashboardEditLayer.tsx:637-640 is undefined for any edit.kind === 'correct' and never checks edit.scalar. The normal menu does check it (:651-653), and Save at :1110 is gated only on blocked.
The fix is very cheap. It adds one edit.scalar check to the existing ternary at :638, reuses the no_correct_shape string that already exists (:653), and needs no new state or mechanism. The cost does not clearly exceed even a LOW harm, so disproportionate-remedy does not apply.

[ADJUDICATION] 81c48b84a25cae130bc794595cf4fe8b204c7b61 total=1 uphold=1 downgrade=0
UPHOLD F1 website/src/pages/members/CrewDashboardEditLayer.tsx:638 reason=harm-warrants-remedy
[GPT-ADJUDICATED] 81c48b84a25cae130bc794595cf4fe8b204c7b61

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 81c48b84a25cae130bc794595cf4fe8b204c7b61 false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 81c48b84a25cae130bc794595cf4fe8b204c7b61 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 81c48b8

Verdict parsed from the review's SHA-scoped output markers for commit 81c48b84a25cae130bc794595cf4fe8b204c7b61.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 81c48b84a25cae130bc794595cf4fe8b204c7b61 false-positive: <reason, 20 to 500 characters>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed readiness: checking Automated validation is still running labels Nov 8, 2026
@chenmingwei23
chenmingwei23 force-pushed the feat/dashboard-edit-mode-lineage branch from ca18617 to bf5480b Compare November 8, 2026 16:19
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Nov 8, 2026
@chenmingwei23
chenmingwei23 force-pushed the feat/dashboard-edit-mode-lineage branch 3 times, most recently from 8aa206e to fbed84b Compare November 8, 2026 16:56
@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Nov 8, 2026
@chenmingwei23
chenmingwei23 force-pushed the feat/dashboard-edit-mode-lineage branch 2 times, most recently from 1891bba to 10d6fd3 Compare November 8, 2026 17:20
@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

Sensitive Change Review

Sensitive files changed:

  • src/kiro_crew/security_posture_sinks_delivery.py

  • @chenmingwei23 posted a complete reasoning template for 81c48b8.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 10d6fd3
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py (50 passed) covers the correction and lineage routes that call the redactor.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, serially: the security posture sink/omission tests above, test_dashboard_edit_mode.py, test_dashboard_stale_fields.py, test_feature_previews_config.py, test_config_baseline.py. CI's Static Ratchet Gate and Backend Lint passed on this head.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 8, 2026
@chenmingwei23
chenmingwei23 force-pushed the feat/dashboard-edit-mode-lineage branch from 10d6fd3 to ce4faf6 Compare November 8, 2026 18:08
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Nov 8, 2026
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: ce4faf6
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py (50 passed) covers the correction and lineage routes that call the redactor.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, serially: the security posture sink/omission tests above, test_dashboard_edit_mode.py, test_dashboard_stale_fields.py, test_feature_previews_config.py, test_config_baseline.py. CI's Static Ratchet Gate and Backend Lint passed on the previous head (10d6fd3); this push changes no line of the sensitive file.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 8, 2026
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 65d97b7e48eb5ac13f0b9eb7bd1b7aacc84e7b1d
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py (50 passed) covers the correction and lineage routes that call the redactor.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, serially: the security posture sink/omission tests above, test_dashboard_edit_mode.py, test_dashboard_stale_fields.py, test_feature_previews_config.py, test_config_baseline.py. CI's Static Ratchet Gate and Backend Lint passed on 10d6fd3 and ce4faf6; the pushes since change no line of the sensitive file (they add tests and facade rows).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23
chenmingwei23 force-pushed the feat/dashboard-edit-mode-lineage branch from ce4faf6 to 6c534ec Compare November 8, 2026 18:45
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 8cb2953
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed on this head) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 4a0aa0f is a rebase onto main a8308a1, which added the chat_runtime_sharing preview. Both previews are kept in previewFlags.ts, FeaturePreviewsSection.tsx and docs/configuration.md; ui-index.generated.json is regenerated, and the Feature Previews counts in previewSurfaces.test.tsx move to 16 switches and 15 hints.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 4a0aa0f.
    • Existing tests (which suites passed on this head): Locally, serially: the security posture tests above, test_feature_previews_config.py (34), test_configuration_doc_facts.py (6), test_lts_feature_previews_off.py (27), test_config_baseline.py (17), the headless boot chain and MCP route table cases, previewSurfaces.test.tsx (37), previewFlagServerSync.test.tsx (23), CrewDynamicDashboard.editGate.test.tsx (9), dashboardEdit.test.ts (61). tsc, eslint, lint:i18n, i18n:check clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 649afa1
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 8cb2953 fixes the GPT 6.1 finding in CrewDashboardEditLayer.tsx: an Escape pressed inside another dialog outside the edit layer (a Modal lets Escape reach window) is left to that dialog and no longer closes the edit popover or drops its draft. New case in CrewDashboardEditLayer.test.tsx fails without the fix; the file has 64 passed, CrewDynamicDashboard.editGate.test.tsx 9 passed.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 8cb2953.
    • Existing tests (which suites passed on this head): Locally, serially: the security posture tests above and the two dashboard test files above. tsc, eslint, lint:i18n and semgrep (CI packs) clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 6870ff7
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed on this head) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 649afa1 is a rebase onto main 3703c2e, which grouped the Feature Previews into cards (feat(settings): group feature previews into cards #35225). The Dashboard edit mode switch now sits in the Chats and Dashboards group in that layout, the feature-map row keeps both main's Reply threads row and this PR's row, and the Feature Previews counts in previewSurfaces.test.tsx move to 17 switches and 16 hints.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 649afa1.
    • Existing tests (which suites passed on this head): Locally, serially: the security posture tests above, test_feature_previews_config.py (34), test_configuration_doc_facts.py (6), test_lts_feature_previews_off.py (27), test_config_baseline.py (17), the headless boot chain and MCP route table cases, previewSurfaces.test.tsx (37), previewFlagServerSync.test.tsx (23), DeveloperPanel.test.tsx (11), CrewDynamicDashboard.editGate.test.tsx (9), CrewDashboardEditLayer.test.tsx (64), dashboardEdit.test.ts (61). tsc, eslint, lint:i18n, i18n:check and docs-lint clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: ce519f0
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" (17 passed) proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 6870ff7 fixes the GPT 6.1 finding in CrewDashboardEditLayer.tsx: the frame's answer to a tap is matched before the stale-hover check, so a pointer move, leave or scroll after the click no longer drops it and the component card opens. New case in CrewDashboardEditLayer.test.tsx fails without the fix; the file has 65 passed, CrewDynamicDashboard.editGate.test.tsx 9 passed.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 6870ff7.
    • Existing tests (which suites passed on this head): Locally, serially: the security posture tests above and the two dashboard test files above. tsc, eslint, lint:i18n and semgrep (CI packs) clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: b101dfd
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since ce519f0 is a clean rebase onto main e8283ce, which carries test(mcp): keep the malformed-type rewriter test independent of PATH #35547 (the fix for the two test_mcp_remote_transport.py TestRewriter failures that main had from fix(mcp): treat a remote entry that carries a command as remote #35335). No PR file changed in the rebase.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to ce519f0.
    • Existing tests (which suites passed on this head): Locally, serially: both TestRewriter cases, the headless boot chain and MCP route table cases, test_feature_previews_config.py (34), test_configuration_doc_facts.py (6), test_config_baseline.py (17), previewSurfaces.test.tsx (37), DeveloperPanel.test.tsx (11), CrewDashboardEditLayer.test.tsx (65), CrewDynamicDashboard.editGate.test.tsx (9). tsc clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 419f1ed
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since b101dfd is a clean rebase onto main e1b079c, which fixes the two main-side test failures from fix(updates): never restart a running gateway; apply at the next start #35088 (test_updates_handlers_facade.py and test_wait_discipline_ratchet.py). No PR file changed in the rebase.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to b101dfd.
    • Existing tests (which suites passed on this head): Locally, serially: the two main-side tests above, the headless boot chain and MCP route table cases, test_feature_previews_config.py (34), test_config_baseline.py (17), previewSurfaces.test.tsx (37), CrewDashboardEditLayer.test.tsx (65), CrewDynamicDashboard.editGate.test.tsx (9). tsc clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 523bd3a
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 419f1ed is a clean rebase onto main 80cf3df, which carries test(updates): cover gateway_restart and update_boot_apply to the 80% floor #35725 (coverage for gateway_restart.py and update_boot_apply.py, the files behind the main-side Coverage Gate red). No PR file changed in the rebase.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 419f1ed.
    • Existing tests (which suites passed on this head): Locally, serially: the headless boot chain and MCP route table cases, test_feature_previews_config.py (34), test_config_baseline.py (17), previewSurfaces.test.tsx (37), CrewDashboardEditLayer.test.tsx (65), CrewDynamicDashboard.editGate.test.tsx (9). tsc clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: f66c597
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 523bd3a is a clean rebase onto main 54b06e2, which carries test(cloud): let the AMI check's fake answer a failed launch's stack lookup #35746 (the fix for the main-side test_cloud_launch_customize.py failure from fix(cloud): say what a failed launch left in aws #35010). No PR file changed in the rebase.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to 523bd3a.
    • Existing tests (which suites passed on this head): Locally, serially: the AMI architecture test above, the headless boot chain and MCP route table cases, test_feature_previews_config.py (34), test_config_baseline.py (17), previewSurfaces.test.tsx (37), CrewDashboardEditLayer.test.tsx (65), CrewDynamicDashboard.editGate.test.tsx (9). tsc clean; gen:settings and gen:ui show no drift.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 770eafb
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" passes on this head (17 passed).
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since f66c597 is a rebase onto main aa89ac8. The only conflict was the generated input_digest in src/kiro_crew/docs/ui-index.generated.json; it was regenerated with npm run gen:ui and the freshness check reports up to date. Every other PR file is byte-identical to f66c597 (git range-diff: commit 1 identical, commit 2 differs only in that generated file).
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to f66c597.
    • Existing tests (which suites passed on this head): Locally, serially: the 6 changed Python test files (323 passed), the 8 changed website test files (1269 passed), test_security_posture.py subset above (17 passed). tsc, eslint on changed files, i18n:check and CI-equivalent semgrep (0 findings) clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 14a4871
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py -k "sink or Omission or redaction_paths" passes (17 passed); the sensitive file is byte-identical to the previously reviewed head.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 770eafb is a rebase onto main b60b6d5. The only conflict was again the generated input_digest in src/kiro_crew/docs/ui-index.generated.json; regenerated with npm run gen:ui, freshness check up to date, gen:settings no drift. Commit 1 is identical by range-diff; commit 2 differs only in that generated file.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. The sensitive file is byte-identical to f66c597 and 770eafb.
    • Existing tests (which suites passed on this head): Locally, serially: the 6 changed Python test files (323 passed), the 8 changed website test files (1269 passed). tsc and i18n:check clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 1d7bd6b
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. The sensitive file is byte-identical to 14a4871.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 14a4871 fixes the two review findings on that head: (1) Opus: _read_lineage passes the live unit only for a package page; a template page passes no unit so read_fields picks the same _write_session unit the template render reads. (2) GPT: the edit layer's open pick and unsent Ask/correction drafts are also kept in a map owned by CrewDynamicDashboard, so a page replacement that unmounts the layer no longer discards them; a draft the person closes is dropped. The redaction call in the lineage module is unchanged.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. _manifest_in_force now returns (manifest, is_package); its only caller and its test are updated.
    • Existing tests (which suites passed on this head): Locally, serially: test/test_dashboard_edit_mode.py (171 passed, incl. a new template-page unit test that fails without the fix), the edit-layer, edit-gate, dashboardEdit, sendChangeRequest and previewSurfaces web tests (178 passed, incl. 3 new draft tests; the remount one fails without the fix). tsc, eslint, flake8, black, mypy, i18n:check, ui-index freshness and CI-equivalent semgrep (0 findings) clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 5de74a3
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. The sensitive file is byte-identical to 1d7bd6b and 14a4871.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 1d7bd6b changes one German string only (website/src/i18n/locales/de.json, dashboardEdit.correct_effect: Zug -> Durchlauf), the fix for the Frontend Tests (4) failure in src/i18n/style/deStyle.test.ts on 14a4871. 1d7bd6b carried the fixes for the GPT 6.1 and Opus 5.5 findings on 14a4871, described in the previous template.
    • Backward compatible (old config / old data / old callers still work, how checked): One translated string; no config, data or caller reads a changed value.
    • Existing tests (which suites passed on this head): Locally, serially: all 13 src/i18n/style/ and destructiveConfirm test files (155 passed; deStyle fails without the string fix), i18n:check clean. Earlier on this tree: test/test_dashboard_edit_mode.py (171 passed) and the edit-layer web tests (178 passed).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: eddabd8
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. The sensitive file is byte-identical to 5de74a3.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. The push since 5de74a3 is a rebase onto main e7e2146 (conflicts: the generated ui-index digest, regenerated; the MCP route-table pin, now main's 351 rows plus this PR's 3 = 354) plus fixes for the three reds on 5de74a3: (1) Opus: the four edit-layer error notices no longer offer the Ask-the-agent hand-off, which would navigate away from the Members page's unsent DM, Profile and crew-editor drafts; (2) GPT: each kept draft value carries an owner token, so a late reply from a closed draft cannot overwrite a draft reopened on the same field; (3) Frontend Ratchets & i18n: the drafts map key is built from dashboardQueryKey instead of a template literal.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value.
    • Existing tests (which suites passed on this head): Locally, serially: the 6 changed Python test files (324 passed), 20 changed/affected web test files (1401 passed, incl. a late-reply test and the no-hand-off assertions, each red before its fix). tsc, eslint on changed files, i18n:check, the i18n strings gate against the base (0 added literals) and ui-index freshness clean.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Re: review at e33377b, the three regex literals in website/src/pages/members/dashboardEdit.ts.

Fixed in 064120f3211445add20fd3a8b7f4869d78c57af3 (the branch is now two commits on current main):

  • parseCorrection boolean branch: const v = raw.trim().toLowerCase(), then ['true', 'yes', '1'] and ['false', 'no', '0'] membership checks (line 232).
  • The path check: spec.startsWith('/') || spec.startsWith('\\') or a drive prefix tested by length, spec[1] === ':', a slash at spec[2] and an ASCII letter at spec[0] (lines 275-276).

No regex literal remains in dashboardEdit.ts, and none was added in the Python modules of this change.

This push also rebases onto main (conflicts were in the preview flag list, the Settings previews list, the settings and UI index generated files, the 13 locale catalogs, and two pinned-digest tests). The generated files were regenerated, and the route-table and entry-types digests were re-pinned to the new values.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 064120f3211445add20fd3a8b7f4869d78c57af3
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 18e6f0405a65e6fc61b2ed8bc4a08f7dfcafbf42
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Disposition for GPT 6.1 on 18e6f0405a65e6fc61b2ed8bc4a08f7dfcafbf42, both fixed in c83b9d0a5e35669d03dcc55ecbfca7bc2abed935:

  • F1 (opening a staged-page link discards unsent drafts): fixed. Edit's on/off state and the drafts map now live in a per-Bot session (useEditSession) held by CrewDashboardTab, which outlasts the keyed dashboard document, and are passed down as editSession. A dashboard rendered elsewhere without it keeps its own session as before. New test keeps an unsent change request when a staged-page link comes and goes (CrewDashboardTab.editDrafts.test.tsx) fails with the fix removed (the typed request is gone after the live page returns) and passes with it.
  • F2 (a late completion overwrites newer text in a restored draft): fixed. Each kept value now carries a revision (useKeptIn in CrewDashboardEditLayer.tsx) that every live write raises. A write from an unmounted box lands only when the revision is still the one its own last write left, so text typed in the remounted box since is never overwritten. New tests keeps text typed after a remount when the earlier send answers late and keeps a correction typed after a remount when the earlier save answers late both fail with the fix removed and pass with it; the file's 71 tests pass.

The branch is still two commits.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: c83b9d0a5e35669d03dcc55ecbfca7bc2abed935
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 29ee9d131dafe0762606aa18db72fb4a22501235
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: fd9749a1f9e6df95828238f8fedd9bac8323f879
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: c2bdab7cac605704792c6eb95b9864db894a74b3
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture*.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (diff is one added tuple).
    • Existing tests (which suites passed on this head): Locally, after rebasing onto current main: test_security_posture*.py, test_dashboard_edit_mode.py, test_feature_previews_config.py, test_config_baseline.py, test_member_dashboard*.py and the facade and route-table pins (1,338 passed in total).
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@KiroCrewCaptain KiroCrewCaptain left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for working through every round on this. One thing still blocks merge on this head.

Where the earlier items stand:

  • Feature-map row says Inspect (docs/feature-map/README.md:212): fixed.
  • _READING_INT_LIMIT is in the crew_log/projection.py manifest (line 183): fixed.
  • The orphan context note for the renamed change keys: fixed.
  • Escape and E while the panel is collapsed: fixed. hiddenFromView guards both listeners (CrewDashboardEditLayer.tsx:369 and :1181).
  • The regexes named in the change request from bolichen97: gone on this head. That review needs a re-review from them.
  1. The UX review could not see the correction flow on this head. Its verdict on c2bdab7 reads PASS, but its body opens "Cannot evaluate: missing the screen after 'Save correction'". It was filed as PASS only because it had no Trigger or Outcome lines. A verdict that cannot be reached for missing evidence counts as a BLOCK. Correcting a value is one of the three goals. The description's correction stills stop at the filled form, and the corrected-value redraw is "left for the narrated end-to-end video recorded in a pod". The description does not link that video. The demo video in the comments was recorded on 3a6ee34, many pushes before this head. Your comment on 377efe8 also says the after-save still is not in yet. These surfaces at head show up in no attachment:

    • the tile and card after a successful save, with summary_person "Set by you ..." (CrewDashboardEditLayer.tsx:1061, :764)
    • the "Use this page" button and its hint (:1107), and paused_newer_page (:637)
    • the failure notices: send_failed and send_unconfirmed (:995, :999), correct_failed (:1100), trace_failed (:1251) and the refused state

    Smallest fix: take stills of those states from this head and add them to the description. Your stubbed Playwright harness can do it with a stubbed correct route and a lineage reply that has actor: "person". The after-save state and the "Use this page" control matter most. The UX lane also says shots 04, 28 and 32 show older wording, so re-shoot those against this head. Then let the UX lane run again.

Separately, the linked issue #33255 is still parked. A maintainer has to sign off the design before this can merge, even after the evidence is in.

Optional:

  • The owning specs still describe the old shapes. learn-cron-dashboard.md says the record body is read for type and data only. crew-log-projection.md lists the custom and agentic rows without source, points, provenance, actor or prior. The two new owner routes are in no spec.
  • crew_log/entry_types_surfaces.py:221 keeps 120 points, but the only reader draws the last 40 values and drops the timestamps. The custom row charge in crew_log/projection_budget.py is sized for 120 and applies with the preview off. A limit of 40 and a measured charge would make that change smaller.
  • dashboard/handlers/member_dashboard_lineage.py:544: a timeout on the instance write lock raises InstanceError, and nothing catches it, so the route returns a bare 500. Returning a coded refusal such as page_busy would match the other refusals on this route.

…e preview

Component ids on view.blocks (data-dashboard-block, checked at the write
gate), the frame's edit bridge and kirocrew.pick, the lineage route, the
person's correction of an agentic value (actor person, told to the Bot on
its next turn), and a script record's declared source/points beside the
gateway's job/script/sha stamp. Preview off composes and folds exactly as
before.
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 17b1d7adb90c03eaf91184caff8a49019b70c501
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (the diff to this file is one added tuple, 12 lines).
    • Existing tests (which suites passed on this head): Run locally after the rebase onto main bffb0b6eda: the 29 Python files this change touches plus test_security_posture.py, test_config_baseline.py, test_feature_previews_config.py, test_facade_mirror_guards.py and test_conductor_agent.py (579 passed in total), and black, isort, flake8 and mypy on the touched Python files. CI results for this head are not claimed here.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Disposition for Opus 5.5 on 17b1d7adb90c03eaf91184caff8a49019b70c501 (member_dashboard.py over the 2,000-line cap), fixed in 1bf9053dbce53f07eb572c6c1ff1aac6fe01e52d:

  • _bound_package_slug, _package_in_force, _package_read and _page_revision moved as whole definitions into the new sibling member_dashboard_package.py, and member_dashboard.py imports them back under the same names, so callers and tests that patch them through member_dashboard are unchanged. The file is now 1,974 lines. No row was added to the line-cap baseline; check_file_line_cap.py reports 0 failures.
  • The pin in test_member_dashboard_package_route.py that lists the files naming the binding wrapper now lists the new module as well.
  • The same push keeps kiro_crew.dashboard_edit_mode out of the MCP listing path: the block attribute constant now lives in dashboard_frame.py (imported by dashboard_edit_mode), and compose imports dashboard_edit_mode where it is used. The listing import budget is back to its main values (485 and 351 modules) with the ceilings unchanged.

…view

Edit toggle on a Bot's Dashboard tab, the capture layer over the sandboxed
frame, the lineage popover (connector line, hop-by-hop, Esc, never over
the value), the component flip with provider and a change request sent
as the person's message, popover Ask/Correct, Sources, the preview card,
the strings in every catalog, the feature map row, the config reference
and the generated indexes.

Part of #33255
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Disposition for GPT 6.1 on 1bf9053dbce53f07eb572c6c1ff1aac6fe01e52d (member_dashboard_lineage.py:468), fixed in 81c48b84a25cae130bc794595cf4fe8b204c7b61:

  • A page whose record holds no apply counter (an older, unpinned record) names itself by the package version, so a package deleted and saved again under the same slug reaches the same page name. read_package_model now marks such a read unpinned (dashboard_package.py), and POST .../dashboard/correct refuses it with 409 page_not_pinned ("apply the dashboard package first") before the stale-page comparison, so nothing is written. A page that has been applied carries a counter that only goes up, and is unaffected.
  • New test test_a_correction_on_an_unpinned_page_is_refused_across_a_delete_and_recreate saves a package with a person-cut version into a real artifact store, deletes it, saves it again, shows the replacement answers to the same page name and version, and then posts the correction. It fails on 1bf9053dbc (200, the entry is written) and passes with the fix.
  • No generation tracking was added. The feature stays behind the default-off preview.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Sensitive change review

  • Head commit: 81c48b84a25cae130bc794595cf4fe8b204c7b61
  • What rule changed: No rule changed. security_posture_sinks_delivery.py gains one row in the redaction-sink registry naming dashboard/handlers/member_dashboard_lineage.py, the new module that scrubs a person's correction reason (redact_via_context) before it enters the Bot's crew log, and returns lineage values through the page's own redaction.
  • Before -> after (what was allowed/blocked, now): Before: the new module called a redactor but was in neither the sink registry nor NON_EGRESS_REDACTION_MODULES, so test_every_redactor_call_site_is_a_registered_sink_or_allowlisted failed. After: it is a registered sink and the Security posture panel counts it. No redactor, gate or allow/deny decision changes.
  • Worst case if wrong (breaks commands / leaks secret / locks user out): The panel's covered-egress count is off by one row. Nothing is blocked or allowed differently; the scrub in the module runs either way.
  • Evidence checked (which sandbox result proves it): N/A, no sandbox rule is touched; the change is a registry row. test/test_security_posture.py passes on this head and proves the row is well-formed and the omission check is satisfied.
  • No regression:
    • Tools still work (which tools/commands were run, result): No tool path changes. test/test_dashboard_edit_mode.py covers the correction and lineage routes that call the redactor and passes on this head.
    • Backward compatible (old config / old data / old callers still work, how checked): Additive row only; no config, data or caller reads a changed value. Every existing sink row is untouched (the diff to this file is one added tuple, 12 lines).
    • Existing tests (which suites passed on this head): Run locally on this head: the Python test files that mention the dashboard package, edit mode or member dashboard, plus the MCP listing import budget and facade mirror guards (2,153 passed), and black, isort, flake8 and mypy on the touched Python files. CI results for this head are not claimed here.
  • How to undo: Revert the one added tuple in src/kiro_crew/security_posture_sinks_delivery.py, and add the module to NON_EGRESS_REDACTION_MODULES with a reason instead.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Converting this PR to draft and pausing pushes. GPT 6.1 raised one new blocking finding on 81c48b84a25cae130bc794595cf4fe8b204c7b61: website/src/pages/members/CrewDashboardEditLayer.tsx:638, where a retained correction stays enabled after the field becomes a list or object (the route then refuses it with not_scalar; nothing is written and the draft is kept). The suggested fix is one edit.scalar check in the existing ternary, reusing the no_correct_shape text. It is not applied yet; a maintainer decision on whether to take it here is pending.

@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Closing for now. The team is reconsidering how much to invest in Bot dashboards versus the conductor-mode Board panel, and the last push (81c48b8) drew another blocker after the declared final push (GPT 6.1 at CrewDashboardEditLayer.tsx:638, retained correction stays enabled after a field turns non-scalar). The branch stays; reopen if the dashboard direction is confirmed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

merge conflict Branch has merge conflicts with its base — author must resolve before merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants