Skip to content

docs: decide goal blocker lifecycle - #13013

Merged
iamwhatever merged 1 commit into
kirodotdev:mainfrom
rubencu:docs/goal-blocker-lifecycle-rfc
Nov 12, 2026
Merged

iamwhatever merged 1 commit into
kirodotdev:mainfrom
rubencu:docs/goal-blocker-lifecycle-rfc

Conversation

@rubencu

@rubencu rubencu commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Record, as one draft RFC that builds on rfc-goal-loop-approval-hold, the goal-loop decisions that design leaves open: remediation contracts, an authority ceiling on the loop's own budget with its sealed commitment record, an owner's ending that automation cannot undo, and finite dashboard goal defaults.

rfc-goal-loop-approval-hold.md (#17033) and its implementation #16993 decided the runtime's response to an unanswered approval: the loop holds. They do not decide how agents treat remediable blockers, who may raise or renew a loop's budget, whether an owner's ending survives a workflow run or cron the loop launched, or what a fresh dashboard goal's budget is. On main the agent-facing contracts still end a goal on a generic blocker, a loop's own session can raise its bounds, stop and re-arm for a fresh budget, or replace its row through ctx.nudge, and a fresh popover goal arms unbounded.

Why it matters

The budget is the service-enforced ending left to a loop that keeps working while an approval holds nothing else back, and today the loop can renew it from its own turn or through automation it scheduled. Agents also keep retiring goals on blockers they could repair, and a dashboard goal nobody budgeted runs until a person stops it.

Not a goal

  • The response to an unanswered approval, the work-ledger bound extension and the person-wait hold: decided by rfc-goal-loop-approval-hold and rfc-work-ledger-person-wait-hold, taken as given here.
  • Implementing any runtime behavior: this PR is documentation only.
  • A finite default for ctx.nudge or the Issue Radar crew runtime.
  • Changing typed structured-monitor outcomes.

What changed (motivation → approach → change)

The RFC is reconciled against the merged hold: every part #17033 decided or #16993 made moot is removed, the undecided parts stand alone, and each kept runtime claim was re-read on main. The file goes from 6,120 to 904 lines.

Removed, by the heading it had:

Kept, as the RFC's decisions:

  • §1 Remediable blockers are work: the producers that still name a generic blocker as a stop (the /goal instruction, the base prompt, the autonudge_stop description, babysit, kirocrew-prepare-pr, the self-nudge recipe and scaffold) and the goal-loop skill's stale approval-stall sentence are aligned to the hold.
  • §2 Authority ceiling: a self-session bound write is capped at the committed pair, save one counted renewal on a work-ledger watch (open question 1, answered (b)), a stop the agent can trigger retains the commitment, a re-arm or proxy at a committed slot inherits it, and an arm at a held loop is refused.
  • §3 The commitment record: autonudge-commitments.json under the OS seal, the file-edit gate and a gateway-only writer, one alias-checking reader, record-over-row authority, write ordering and the owner-recommit signal.
  • §4 Commitment generation and turn provenance (Phase 2): the inheritance rule, the sealed provenance triple and definition digest (structured monitors included), the /goal provenance gate, channel provenance, and app watchdogs that honor an ended commitment.
  • §5 Arming surfaces: a 50-cycle default for the popover and POST /api/autonudge, the popover's provenance and draft marker, and negative caps refused everywhere.

New: a "Relationship to rfc-goal-loop-approval-hold" section, a "Relationship to rfc-permanent-patrol" section, and three open questions. Open question 1 records that the §2 cap refuses the monitor_update raise rfc-goal-popover-pause-play-controls describes, and refused the conductor's patrol_budget.py renew, which #17033 kept until #33385 retired it; bolichen97's maintainer decision answers it with (b), so the cap admits a renewal on a work-ledger watch within three renewals, 1,000 cycles and seven days, counted in the sealed record. Main has since retired agent-side renewal (#33385, #33599), so the RFC asks the maintainer to confirm that (b) stands. Open question 2 asks whether the hold's fields move into the sealed record. Open question 3 asks what a self-session write may do to a slot with no readable record: be capped at the row's stored bounds, or leave the gap the Security section states; the Phase 1 cap is blocked on it. The acceptance requested here is a maintainer record that answers open question 3, or splits the Phase 1 bound-write cap out to wait for it (Phase 0).

Front matter follows the directory's conventions: status draft (no maintainer acceptance recorded), doc-pr: 13013, no implementation PRs. The README gains this RFC's index row beside rfc-goal-loop-approval-hold and one audit sentence.

Backwards compatibility

Compatible: documentation only. No runtime behavior, API, stored data or user-visible surface on main changes.

Removes nothing: the diff adds one RFC and one index row to docs/request-for-change/ and deletes no file, record or documented capability on main.

Tests

  • python scripts/docs_lint.py: all documentation checks passed.
  • BRAND_BASE_REF=origin/main python scripts/check_brand_name.py: no misspellings in added lines.
  • COMMENT_HISTORY_BASE_REF=origin/main python scripts/check_comment_history.py: passed.
  • Current head bcfdf4ee09 is one commit directly on 9111b0d2cc. Since 2dccca3b71 it changes the README audit line, where this RFC's sentence stays last on main's line, after every sentence that line carries on 9111b0d2cc (main's line is kept as it is), and the aliased-leaf replacement, which now keeps the cutover marker. No runtime code changed. The RFC is re-audited against ea45e802e9 (last-audited: 2026-10-20, audited-at: ea45e802e9): every source path and code identifier it cites exists there (subject_outcome is the field the open fix(autonudge): read a legacy loop's pull request between cycles and while held #21377 proposes), except the ones the rfc-permanent-patrol re-read below cites on 9111b0d2cc. All 52 repository paths it cites exist on 9111b0d2cc; main changed 48 of them since 2d5bb2f86d, the base of the previous check. Every one of the RFC's 153 backticked identifiers that src/, test/ or website/src/ held on 2d5bb2f86d is still there on 9111b0d2cc, and each of the 31 cited on the same line as a repository path that held there still appears in that file. The six the sweep does not find on either base are names the RFC or an open PR proposes (autonudge-commitments.json, subject_outcome), three written as Class.member, and the audit sha ea45e802e9. Each citation of code main moved since 52a302b914 names its new owner: _handle_goal_command and the /goal instruction (dashboard/chat_turn/slash_commands.py, refactor(dashboard): move the turn queue and eager spawn into chat_turn #21918), the autonudge_stop description (mcp_tools/control_parts/schemas.py, refactor: split mcp_tools control.py and spawn.py into owner packages #21575), the _monitor_update applier (dashboard/session_directive_apply_monitor.py, refactor(dashboard): split change_cards, chat_tags and directive apply #22138), the store's agent-writable statement (autonudge_service/store_loader.py, refactor: split history_projection, snapshot and autonudge into owners #22116), and the Slack cron turn that adds acked_items to the prompt (slack/gateway_runtime/cron_turn.py). The Design round on a738a0cb55 is answered. A new Alternatives entry, "Hold Phase 2 until it also bounds a self-scheduled cron", weighs Phase 2's cost against the cron residual. Phase 0's exit criteria require the accepting maintainer to record that a person's chat stop request is a stop from a turn, retained under §2. Phase 1 lands as two PRs, Phase 1a first and on its own. The Design round on d4879d809d is answered too: Phase 1b, which carries that §2 rule, does not start without the record; acceptance can be recorded in one record or two, so Phase 1a can start once §1, §5 and the negative-cap refusals are accepted; and Phase 2 waits for a separate maintainer decision on its definition-digest cost against the cron residual. The index row now cites ea45e802e9, the commit the audit records, and its six claims hold there and on current main. The review of 16d2df0b40 is answered as well: the Status line at the top now says acceptance comes in one record or two and the status flips to accepted once both are recorded, and it names Phase 1's two PRs; "Hold Phase 2 until it also bounds a self-scheduled cron" now calls its rejection this document's recommendation, with the decision itself the accepting maintainer's under Phase 0. The Opus finding on 0ff5fd4ff7 is answered too: a row written before Phase 1 now has its record sealed from its stored bounds at the first load after the upgrade, so the agent-writable row never decides a later load (Backward compatibility, the cutover exception in §3, and a new Phase 1 exit criterion). bolichen97's two points on 25188f8cb1 are answered as well: the cutover runs only while the sealed leaf lacks its cutover marker, which the gateway writes in the same atomic write as the records it seals, rows or none; and the Phase 1 criterion now also covers an out-of-band max_runtime_secs raise and a row written after the cutover, which stays uncommitted. His follow-on on 2dccca3b71, which the crew-pr-reviewer raised again on 60d2aee7c1, is answered too: the fresh empty record that replaces an aliased leaf (One reader that trusts no name, and the Phase 1 test item) keeps the cutover marker, because a writer runs only after a load that has completed the cutover, so the next load does not run the cutover again and seal whatever bounds the rows hold then. The Opus finding on 5e8f9a4d7f is answered as well: the Security section's "Fail closed" bullet no longer says a self-session write cannot raise a row's bounds when the row's record is missing or aliased. It now states that gap, since such a row is uncommitted and the cap of §2 applies only against a committed pair, and a new open question 3 records the choice between capping such a write at the row's stored bounds and accepting the gap. bolichen97's re-review of b24b20282f is answered too: Phase 0's exit criteria and the Phase 1 paragraph now name open questions 1 and 3 as what the bound-write cap waits on, so the acceptance record cannot close with that choice unrecorded. The crew-pr-reviewer's r146 HOLD on eb2a3c8539 is answered as well: the README conflict is cleared, and bolichen97's decision on open question 1, (b), is recorded in the RFC: under that question, in §2's bound-write rule and its renewal paragraph, in §3's record, in Phase 0, in Phase 1 with a new exit criterion for the counted renewal, and in the Summary, Backward compatibility, Security, the schema-ceiling alternative and open question 3. Main's rfc-permanent-patrol (docs: RFC for loops the model arms picking no cycle cap #33382, implemented by feat: loops the model arms pick no cycle cap and end with one final report #33385) and chore(conductor): remove the unused patrol_budget tool and script #33599 changed the loop defaults and removed agent-side renewal after that head, so this head re-reads those statements on 9111b0d2cc. That covers a new "Relationship to rfc-permanent-patrol" section, the header, the Motivation's monitor_update and monitor_start bullets, and §2's dispatch patrol and renewal paragraph. It also covers the §5 rows for monitor_start and the dispatch patrol, the "Why 50" paragraph, the Phase 1 renewal criterion, Security's no-new-authority bullet, two Alternatives, and one line under open question 1's recorded answer. The design of §2 to §4 and the status are unchanged, and the rest of the audit stands at ea45e802e9. Earlier rounds' changes are unchanged: the autonudge.json sealing alternative, the definition digest's cost, the autonudge_selfarm.py precedent, the dispatch patrol arm (§2, §5, Phase 1 and compatibility), the runtime budget's clock, the stop/re-arm/restore renewal fix, the self-scheduled cron residual, the time-window alternative to Phase 2, and the Phase 1a/1b split. The README index row is rebased onto main's rows on 9111b0d2cc, which are kept as they are, as is main's audit line: the README differs from main only by this RFC's row and its one audit sentence, which names the audit sha. Documentation lint, comment-history, brand-name, harness-parity, diff, the flake-pattern and wait-discipline ratchets, and the single-commit push guard pass.
  • git diff --check origin/main..HEAD: clean.
  • push_guard.py --base main --require-single-on-base: STATUS: SAFE TO PUSH (single commit on base).
  • Each runtime claim the RFC keeps was re-read on origin/main and cites the file it describes.

Manual verification

N/A — documentation only; the factual check against main is listed under Tests.

Related Issues

Builds on #17033 and #16993.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • 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

Contribution License Agreement

Contributed under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

@rubencu
rubencu requested review from a team and bolichen97 September 23, 2026 06:00
@github-actions github-actions Bot added the fork Pull request from a fork (external contributor) label Sep 23, 2026
@rubencu

rubencu commented Sep 23, 2026

Copy link
Copy Markdown
Contributor Author

Intent: Record the maintainer decision that remediable blockers and human-only approvals keep autonomous goals active within finite budgets and a strict no-self-grant authority ceiling.
Not a goal: Implement runtime behavior, remove explicit stop controls, or change typed structured-monitor delivery-failure semantics.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Sep 23, 2026
@github-actions

github-actions Bot commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — ✅ no blocking findings

Reviewed bcfdf4ee09319dc8bd6932b1da0d3aae10fcc827 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] bcfdf4e

@github-actions

github-actions Bot commented Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — ✅ PASS

Design-level review of bcfdf4ee09319dc8bd6932b1da0d3aae10fcc827 via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks: this is a draft RFC with no runtime change, and it splits each costly phase (1b, 2) behind its own recorded maintainer decision.

Notes

  • The description says the RFC "goes from 6,120 to 904 lines", but the patch adds a 1,183-line file. Correct the count so reviewers can tell which revision was reconciled.
  • §4's watchdog rule edits issue_radar/backend/crew_runtime.py and auto_research/campaign/watchdog.py in core, but both apps are leaving under Move every built-in app out to KiroCrewApps and forbid built-in apps #26967. Write it as a generic App Kit seam (any app reads "is this slot's commitment open"), so Phase 2 needs no app-specific code in core.
  • Open question 1's answer (b) now admits a renewal that nothing on main calls (chore(conductor): remove the unused patrol_budget tool and script #33599 removed it). Closing it as "no exception" would drop the renewal counter from the sealed record and the Phase 1b tests, which makes Phase 1b simpler.

[DESIGN-REVIEWED] bcfdf4e

Description read

[DESCRIPTION-READ] b9a316d0cdd917f4bf723555cf64c8d6d915cfa64c10dbe0ee684d09bcf60aaf

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 Sep 23, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — 🔴 changes requested (blocking)

Reviewed bcfdf4ee09319dc8bd6932b1da0d3aae10fcc827 via the fork AI-review pipeline; updated in place on each push.

The new rfc-goal-blocker-lifecycle row in the RFC index says main does something it doesn't. It claims the dispatch patrol re-arms with a 300-cycle, 86,400-second pair, but main sets no cycle cap and uses the operator's runtime ceiling. The RFC's own §5 table, in this same diff, says the opposite.

BLOCKING — docs/request-for-change/README.md:89

the dispatch patrol (conductor_patrol.ensure_patrol) replaces a STOP-sentinel stop with a fresh 300-cycle, 86,400-second pair
Trigger: Someone reads the "What is actually on main" column for this RFC, as the First Principles lane does. ensure_patrol in src/kiro_crew/conductor_patrol.py:407-414 arms with max_cycles=PATROL_MAX_CYCLES (= 0, line 52) and max_runtime_secs=patrol_max_runtime_secs(), which returns the operator's monitoring.max_runtime_secs ceiling. The 300 is PATROL_INTERVAL_SECS, the idle interval, not a cycle count.
Outcome: The index tells readers that a re-armed patrol stops after 300 cycles or one day. On main it has no cycle cap and runs until the operator ceiling (7 days as shipped). That contradicts the RFC's own §5 row ("PATROL_MAX_CYCLES = 0, no cap … patrol_max_runtime_secs, the operator ceiling").
Anchor: defect
Fix: Reword the clause to say the patrol replaces a STOP-sentinel stop with a fresh arm that has no cycle cap (PATROL_MAX_CYCLES = 0) and the operator runtime ceiling (patrol_max_runtime_secs).

[BLOCK-MERGE] bcfdf4e
[OPUS-REVIEWED] bcfdf4e

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running and removed readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention labels Sep 23, 2026
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Sep 23, 2026
@rubencu
rubencu force-pushed the docs/goal-blocker-lifecycle-rfc branch from 795067b to 671cf22 Compare September 24, 2026 20:15
@github-actions github-actions Bot added readiness: checking Automated validation is still running 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: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Sep 24, 2026
@rubencu
rubencu force-pushed the docs/goal-blocker-lifecycle-rfc branch from 671cf22 to f63249a Compare September 25, 2026 00:52
@rubencu

rubencu commented Sep 25, 2026

Copy link
Copy Markdown
Contributor Author

span=ab637e60142f

  • Unbounded-loop lifecycle contradiction

Fixed in f63249a.
Section 5 now defines one permanent per-loop rule: any non-zero cycle or runtime bound continues remediation; both bounds at 0 retain the terminal approval_stalled stop.
The migration and Phase 1 exit criteria explicitly keep pre-upgrade stored 0 values unchanged and on the unbounded branch.

@rubencu

rubencu commented Sep 25, 2026

Copy link
Copy Markdown
Contributor Author

span=14be823d2b60

  • Unbounded programmatic arming surfaces

Fixed in f63249a.
New section 6 inventories ctx.nudge, Issue Radar, and auto-research alongside every other direct arming and re-arm path.
Those caller-owned contracts are explicitly excluded from Phase 1 defaults and retain the approval_stalled stop whenever both stored bounds are 0.

@rubencu

rubencu commented Sep 25, 2026

Copy link
Copy Markdown
Contributor Author

span=14be823d2b60

  • Measured monitor_start defaults

Fixed in f63249a.
The measured base now names the existing 24-cycle and 14,400-second monitor_start defaults and its schema-level rejection of 0.
Section 6 justifies the dashboard goal value of 50 from shipped /goal parity and the shorter goal cadence, while keeping runtime opt-in for those surfaces.

@rubencu

rubencu commented Oct 22, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main a932db73bb to clear the README conflict: ce2da823f0 -> 3efdede0f6. Main's #21981 appended its rfc-remote-computer-use sentence to the same audit line, and #22983 added a row. Main's line is kept as it is, and this RFC's sentence stays last on it. This RFC's row sits after main's rows in the table. Checked: the resolved README equals main's file plus exactly this PR's sentence and row, and the RFC file is unchanged (same patch ID). Docs tests, the script gates, the ratchets and the push guard pass.

ce2da823f0's approved round had finished green before the conflict appeared. The runs on 3efdede0f6 wait at action_required for a maintainer.

@rubencu

rubencu commented Oct 22, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 7d9673ba7b as 5e8f9a4d7f, because the PR had become conflicting in docs/request-for-change/README.md. Main's #22323 added an audit sentence and a row, and #23039 and #23049 added rows. Main's audit line is kept as it is with this RFC's sentence last, and this RFC's row sits beside theirs: the resolved README equals main's plus exactly this PR's sentence and row. The RFC file itself is patch-identical.

Checked on this head: the documentation tests that read the RFC index (1,094 passed), docs lint and the other script gates, the flake-pattern and wait-discipline ratchets, and the single-commit push guard. The body's current-head paragraph is regenerated: all 50 repository paths the RFC cites exist on this main, and the identifiers it cites beside the six cited files main changed since a932db73bb are still there, directly or through their facades.

@rubencu

rubencu commented Oct 22, 2026

Copy link
Copy Markdown
Contributor Author

This is the round push for 5e8f9a4d7f: b24b20282f on main 69b2e503eb. It answers the Opus 5.5 finding on 5e8f9a4d7f (comment 5790035564). The RFC's Security section said a missing or aliased record still keeps a self-session write from raising a row's stored bounds. No rule provided that: such a row is uncommitted, and the cap of §2 applies only against a committed pair. The "Fail closed" bullet now states the gap as it is. A new open question 3 records the choice for the accepting maintainer: (a) cap a self-session bound write at the row's stored bounds whenever its slot holds no readable record, or (b) accept the gap, because removing or aliasing a record takes a write to the protected leaf from outside the default seal. The design itself is unchanged.

The README stays main's plus this RFC's sentence and row. Checked on this head: the documentation tests that read the RFC index (1,094 passed), docs lint and the other script gates, the flake-pattern and wait-discipline ratchets, and the single-commit push guard. The body's current-head paragraph is regenerated.

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Re-reviewed the change since 5e8f9a4 on head b24b202. The head was rebased onto main; the README edits are unchanged, and the RFC changed in two places.

  • The "Fail closed" bullet no longer overclaims. It now says plainly that a missing or aliased record leaves the row uncommitted, so the cap of section 2 does not apply and the loop's own session can raise that row's bounds until an owner recommits. That matches section 3, which decides a failed leaf check as no record. Good correction.
  • New open question 3 lays out the two choices for that gap. Option (b) rests on the default agent.sandbox of "auto"; that default matches main.

One small gap: open question 3 says the Phase 1 cap "already waits on open question 1, and this choice joins it", but the Phase 0 exit criteria and the Phase 1 paragraph still name only open question 1 as what the bound-write cap waits on. Suggest adding open question 3 to both, so the acceptance record cannot close with this choice unrecorded.

Design sign-off on this RFC stays with the maintainers; this review does not approve it.

@rubencu

rubencu commented Oct 22, 2026

Copy link
Copy Markdown
Contributor Author

This is the round push for b24b20282f: f547c9cc30 on main 0be5bcd9ea. It answers bolichen97's re-review of b24b20282f (review 5618733628). Phase 0's exit criteria and the Phase 1 paragraph now name open questions 1 and 3 as what the bound-write cap waits on, so the acceptance record cannot close with the open question 3 choice unrecorded. Nothing else in the RFC changed, and the README is patch-identical.

The other reds on b24b20282f are not from this diff, which touches only the RFC and the README index. Backend shard 6 hit a 30-second subprocess timeout in test_push_guard_squash.py's SIGINT case. The offline E2E failed because one sidebar-tags spec passed only on a retry, and the Coverage Gate failed closed after that shard. Security Scope never resolved its review scope. Opus 5.5 and GPT 6.1 report no blocking findings there, and Design passed.

Checked on this head: the documentation tests that read the RFC index (1,094 passed), docs lint and the other script gates, the flake-pattern and wait-discipline ratchets, and the single-commit push guard. The body's current-head paragraph is regenerated.

@bolichen97 bolichen97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Re-reviewed the change since b24b202 on head f547c9c. The head was rebased onto main; the README edits and the commit message are unchanged, and the RFC changed in two lines.

  • The Phase 0 exit criteria now require answers to open questions 1 and 3, or the Phase 1 bound-write cap is split out to wait for them.
  • The Phase 1 paragraph now says the bound-write cap is blocked on open questions 1 and 3.

That closes the gap raised in the previous review: the acceptance record can no longer close with the open question 3 choice unrecorded. No new findings.

Design sign-off on this RFC stays with the maintainers; this review does not approve it.

@rubencu

rubencu commented Oct 22, 2026

Copy link
Copy Markdown
Contributor Author

Round push for f547c9cc30: c301cd0630 on main 840dc28664, a rebase only.

f547c9cc30's round finished clean. CI and PR Readiness passed, GPT 6.1 and Opus 5.5 report no blocking findings, Design passed, and First Principles and UX were skipped. bolichen97's re-review of that head (5621002441) found no new findings.

Main's #23166 then added an audit sentence and an index row to docs/request-for-change/README.md, which conflicted with this RFC's audit sentence. Main's line and rows are kept as they are, with this RFC's sentence last on the audit line and its row where it was. #21746's change to the crewmate-share row merged cleanly. The RFC itself is byte-identical to f547c9cc30. Of the repository paths it cites, main changed dashboard/chat_runner.py (#22914) and workflows/service.py (#23124) since 0be5bcd9ea. Every identifier the RFC cites beside them (_run_chat, _directive_user_origin, _nudge_port, authorize_and_add_nudge) is still there.

Checked on fc659a95ea, the same patch on 66b3698653: the documentation tests that read the RFC index (1,094 passed), docs lint and the other script gates, the flake-pattern and wait-discipline ratchets, and the single-commit push guard. Main's seven newer commits touch none of this PR's files, and the push guard passes again on c301cd0630. The body's current-head paragraph and citation counts are regenerated. A recount this pass finds 51 cited paths, where the body said 50.

@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD at c301cd0.

  1. The branch conflicts with main again, in docs/request-for-change/README.md. A rebase clears this.
  2. The RFC is still draft, and its own Phase 0 (L677-L698) requires a maintainer acceptance record. That record has to answer open questions 1 and 3 (L1077-L1108), or the Phase 1 bound-write cap has to be split out. Only a maintainer can clear this.
  3. The Design note on this head is unanswered. The description says "two open questions" but the file has three. Please update the description so the acceptance request names question 3.

The aliased-leaf item from r117 is resolved: the replacement record keeps the cutover marker (L385-L389). Our 2026-10-11 change request is out of date because its defect was fixed, and it can be dismissed. The AI lanes report no blocking findings, and the checks are green.

— crew-pr-reviewer (r122)

@rubencu

rubencu commented Oct 23, 2026

Copy link
Copy Markdown
Contributor Author

Rebase for the README conflict: c301cd0630 -> cc3a243154 on main 2d5bb2f86d.

Main's #23208 added an audit sentence and an index row to docs/request-for-change/README.md, which conflicted with this RFC's audit sentence again. Main's line and rows are kept as they are, with this RFC's sentence last and its row where it was. The RFC file itself is unchanged.

crew-pr-reviewer r122 item 3 and the Design note on c301cd0630: the description said the RFC adds two open questions, but the file has three. The What changed section now names open question 3. It also says the acceptance requested is a maintainer record that answers open questions 1 and 3, or splits the Phase 1 bound-write cap out to wait for them, as Phase 0 says. The Design note's second point is to keep Phase 1b's cap out until questions 1 and 3 are answered, not bundle it with them. That is what Phase 0's exit criteria and the Phase 1 paragraph already require, so the RFC is unchanged there. Item 2, the acceptance record itself, is a maintainer's.

Of the 51 repository paths the RFC cites, main changed four more since 840dc28664: dashboard/chat_api/slot_lifecycle.py, dashboard/chat_runner.py, dashboard/chat_turn/slash_commands.py and website/src/components/AutoNudgePopover.tsx. Every identifier the RFC cites beside them is still there: _retire_slot_nudge_loop, _restore_slot_nudge_loop, _run_chat's _directive_user_origin, _handle_goal_command with its "Hard blocker" line, and parseCycles.

Checked on this head: the documentation tests that read the RFC index (1,098 passed), docs lint and the other script gates, the flake-pattern and wait-discipline ratchets, and the single-commit push guard. The body's current-head paragraph and citation counts are regenerated.

@rubencu

rubencu commented Oct 29, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 14ef5d7307 to clear a conflict in docs/request-for-change/README.md: main had added more sentences to the index audit line. Main's line is kept as it is, with this RFC's audit sentence still last; the README now differs from main only by this RFC's row and that one sentence. The RFC itself is unchanged.

All 51 repository paths the RFC cites still exist on 14ef5d7307, and every code identifier it cites that was in the tree on 2d5bb2f86d, the previous check's base, is still there. Docs lint, the docs and RFC tests (1,613 passed), the script gates, the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name 82c2242021.

@rubencu

rubencu commented Oct 29, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 0f277f1511 to clear another conflict in docs/request-for-change/README.md: main's #25380 changed the index audit line. Main's line is kept as it is, with this RFC's audit sentence still last; the README differs from main only by this RFC's row and that one sentence, and the RFC is unchanged.

All 51 repository paths the RFC cites still exist on 0f277f1511, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (1,621 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name 73344648ae.

@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD at 046c21e.

  1. Merge conflict. The branch conflicts with main again (CONFLICTING). As in earlier rounds, the likely file is docs/request-for-change/README.md. A rebase clears this.
  2. Maintainer decision. The RFC is still draft, and its Phase 0 requires a maintainer acceptance record that answers open questions 1 and 3, or splits the Phase 1 bound-write cap out. Only a maintainer can clear this.

The diff is documentation only. The AI lanes report no blocking findings. Design's two notes on this head are guidance for that acceptance record and do not block. Our 2026-10-11 change request is out of date and can be dismissed.

— crew-pr-reviewer (r135)

@rubencu

rubencu commented Oct 29, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 043dd9fd7f to clear another conflict in docs/request-for-change/README.md: main's #23181 added a sentence to the index audit line. Main's line is kept as it is, with this RFC's audit sentence still last; the README differs from main only by this RFC's row and that one sentence, and the RFC is unchanged. This clears item 1 of the crew-pr-reviewer r135 HOLD; item 2, the maintainer acceptance record, stays a maintainer's.

All 51 repository paths the RFC cites still exist on 043dd9fd7f, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,028 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name 7c39562650.

@bolichen97

Copy link
Copy Markdown
Contributor

This branch conflicts with main again, most likely in docs/request-for-change/README.md, where main has added several RFC rows since this head. Could you rebase onto current main and keep both sets of rows? The design question on this RFC is still open, so this is only about keeping the branch mergeable.

@rubencu

rubencu commented Nov 4, 2026

Copy link
Copy Markdown
Contributor Author

@bolichen97 Rebased onto main 20c5654909, keeping both sets of rows. The only conflict was the index audit line in docs/request-for-change/README.md. Main's rows and audit line are kept as they are, and the README differs from main only by this RFC's row and its one audit sentence. The RFC is unchanged.

All 51 repository paths the RFC cites still exist on 20c5654909, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,522 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name ac50d42717.

@rubencu

rubencu commented Nov 5, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main bd9c87744b to clear another conflict in docs/request-for-change/README.md: main's #24384 changed the index audit line and one row. Main's rows and audit line are kept as they are, with this RFC's audit sentence still last; the README differs from main only by this RFC's row and that one sentence, and the RFC is unchanged.

All 51 repository paths the RFC cites still exist on bd9c87744b, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,524 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name eb2a3c8539.

@bolichen97

Copy link
Copy Markdown
Contributor

@iamwhatever could you re-review? The hole from your change request at 50ca304 looks closed at eb2a3c8. At load, an open record with no row and no pending-transition mark now becomes a retained stop that the next monitor_start from the session inherits, instead of closing. Only a crash inside a transition is resolved from the mark the writer sets before the row. So deleting your own row and waiting for a restart no longer buys a fresh budget (§2, and the bullet near the end of the decisions list).

Open question 1 (the §2 cap against patrol_budget.py renew: option a, b or c) still needs a maintainer pick; that is not something the author can settle.

@bolichen97

Copy link
Copy Markdown
Contributor

Maintainer decision: the direction is signed, including the §2 cap on a goal loop's own budget writes.

For open question 1 we pick (b): the cap admits a renewal on a work-ledger watch within the limits patrol_budget.py already enforces (three renewals, 1,000 cycles, seven days), counted in the sealed record. Why: it is the only option that keeps today's renew working unchanged within those limits, so it refuses nothing that works now; (a) retires renewal and (c) drops the cap.

This is a decision on the RFC's direction, not an approval of the PR.

@rubencu

rubencu commented Nov 5, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the decision. I'll record open question 1 = (b) in the RFC, linked to your decision, in one push after iamwhatever's re-review of eb2a3c8, or sooner if a maintainer prefers. That keeps the head he was asked to review in place.

Per Phase 0's exit criteria, the document still needs:

  • an answer to open question 3;
  • the record that a person asking the agent in chat to stop is a stop from a turn, which §2 retains (Phase 1b waits for it);
  • a separate decision on Phase 2 (Phase 2 does not start without it);
  • the acceptance record, in one record or two (§1, §5 and the negative-cap refusals; §2 to §4).

@iamwhatever

Copy link
Copy Markdown
Contributor

HOLD at eb2a3c8: the branch conflicts with main again, in docs/request-for-change/README.md.

Our r99 point is closed on this head. At load, an open record with no row and no pending-transition mark now becomes a retained stop that the next monitor_start inherits (rfc-goal-blocker-lifecycle.md#L404-L413, #L752-L753). Deleting a row and waiting for a restart no longer buys a fresh budget. Every check on this head is green or skipped.

To clear:

  1. Rebase onto main to clear the README conflict.
  2. In the same push, record @bolichen97's decision on open question 1 (option b) in the RFC, as you planned. That push needs a fresh review, so there is no reason to keep this head in place.

The remaining Phase 0 items you listed (open question 3, the chat-stop record, the Phase 2 decision, the acceptance record) are exit criteria inside the RFC, not blockers for merging a draft. Design sign-off stays with the maintainers.

— crew-pr-reviewer (r146)

@rubencu

rubencu commented Nov 5, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 268a10815a to clear the conflict in docs/request-for-change/README.md: main's #29996 changed the index audit line and one row, and #29145 added a row. Main's rows and audit line are kept as they are, with this RFC's audit sentence still last; the README differs from main only by this RFC's row and that one sentence.

The same push records bolichen97's decision on open question 1, option (b): the cap admits a renewal on a work-ledger watch within the limits patrol_budget.py already enforces, counted in the sealed record. It is recorded under the question and in each passage that states the cap or cites the question, with one new Phase 1 exit criterion for the counted renewal. Status stays draft. Open question 3, the chat-stop record, the Phase 2 decision and the acceptance record are still open.

All 51 repository paths the RFC cites still exist on 268a10815a, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,559 passed), the flake and wait ratchets and the single-commit push guard pass. The description now names 1708020f85 and records the answer too.

@bolichen97

Copy link
Copy Markdown
Contributor

Checked 1708020: open question 1 now records option (b) as decided, with the same limits (three renewals, 1,000 cycles, seven days), counted in the sealed record, and the reason matches the decision. §2, §3 and the new Phase 1 exit criterion for the counted renewal say the same thing. Status stays draft, which is right while open question 3, the chat-stop record, the Phase 2 decision and the acceptance record are still open.

No further change needed from me on this point.

@rubencu

rubencu commented Nov 6, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 35a2472ecd to clear another conflict in docs/request-for-change/README.md: main's #30788 changed the index audit line and added a row. Main's rows and audit line are kept as they are, with this RFC's audit sentence still last; the README differs from main only by this RFC's row and that one sentence, and the RFC is unchanged.

All 51 repository paths the RFC cites still exist on 35a2472ecd, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,561 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name edaab6409c.

…omous work

Builds on rfc-goal-loop-approval-hold, whose hold (kirodotdev#16993) decides the
runtime's response to an unanswered approval, and proposes what that
document does not decide:

- agent-facing contracts that treat a remediable blocker as work rather
  than a stop condition, across the base prompt, the /goal instruction,
  the autonudge_stop description and the bundled and repo skills;
- an authority ceiling that covers the loop's own budget: a self-session
  bound write is capped at the pair the owner committed, save one counted
  renewal on a work-ledger watch, a stop the agent can trigger retains the
  commitment, and a re-arm or proxy at a committed slot inherits it;
- a per-slot commitment record in a leaf the sandbox seals, the file-edit
  gate refuses and only the gateway writes;
- a commitment generation and sealed provenance for scheduled automation,
  so a run, cron or completion launched under an ended commitment arms
  nothing; and
- a finite 50-cycle default for the dashboard goal surfaces and the
  refusal of a negative cap at every boundary.

It keeps rfc-permanent-patrol's decision (kirodotdev#33382, kirodotdev#33385): a loop the
model arms picks no cycle cap, and a work-ledger watch takes no bound from
the model. What the ceiling still bounds for such a loop is its runtime
window, which a self-session stop and re-arm would otherwise restart.

The bound-write cap collided with the conductor's agent-side renewal,
which the approval-hold document kept. Open question 1 records the
maintainer's answer, (b): the cap admits a renewal on a work-ledger watch
within three renewals, 1,000 cycles and seven days, counted in the sealed
record. Main has since retired that renewal (kirodotdev#33385, kirodotdev#33599), so the
answer is left for the maintainer to confirm. The Phase 1 cap still waits
on open question 3.

Adds the document's row to the request-for-change index.
@rubencu

rubencu commented Nov 12, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main 9111b0d2cc to clear another conflict in docs/request-for-change/README.md (main's index audit line changed again). Main's rows and audit line are kept as they are, with this RFC's audit sentence still last.

Since the last head, main accepted rfc-permanent-patrol (#33382, implemented by #33385) and removed the patrol_budget tool and script (#33599). A loop the model arms now picks no cycle cap by default, a work-ledger watch takes no bound from the model, and nothing renews a patrol from the agent side. This head re-reads the RFC's statements about those defaults on 9111b0d2cc and adds a short "Relationship to rfc-permanent-patrol" section. The design of §2 to §4, the recorded answers and the draft status are unchanged, and the index row keeps its claims as verified at ea45e802e9.

One point for bolichen97: open question 1's answer, (b), admits a counted renewal on a work-ledger watch, and main no longer makes that renewal. The RFC keeps (b) as recorded and asks whether it still stands for a renewal a later change might add, or whether the question now closes with no exception.

All 52 repository paths the RFC cites exist on 9111b0d2cc, and no code identifier it cites has gone missing since 2d5bb2f86d. The docs and RFC tests (2,653 passed), the flake and wait ratchets and the single-commit push guard pass. The description's exact-head claims now name bcfdf4ee09.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Code change approver recommendation: A person should decide

Why: the decider thinks the change is probably fine but estimates a 15% chance that this change introduces a defect, above the 12% limit for approving on its own.

This PR changes only docs. It adds a draft RFC (rfc-goal-blocker-lifecycle) plus one index row and one audit sentence in the RFC README, and changes no runtime code. I checked most of the index row's runtime claims against the repo and they hold on current main. The /goal instruction still says autonudge_stop(reason="blocked"). _add_unserialized/_update_unserialized still store max(0, int(max_cycles)). MONITOR_UPDATE_SCHEMA still accepts max_cycles from 1 to 1000. The RFC body's §5 table matches current conductor_patrol.py (PATROL_MAX_CYCLES = 0, patrol_max_runtime_secs). The design tightens a trust boundary: it adds a sealed commitment record and a cap on a loop session writing its own budget. It names its remaining gap (open question 3) and blocks the Phase 1 cap on a maintainer answering it. The status is honestly draft, and the RFC asks for a maintainer acceptance record before anything is built. One claim is out of date: the index row's patrol claim, noted below. The README says each row's audited-at is the freshness signal, so a row dated to a commit fits the directory's convention and does not block. The CI checks that finished passed. The pending one

An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit bcfdf4ee09 · 1186 changed lines. If this recommendation is wrong, reply here: replies are how the approver is evaluated.

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

Labels

fork Pull request from a fork (external contributor) needs-author-decision PR blocked on author input

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants