Skip to content

docs(rfc): peer-sent turns may reset their own conversation - #18338

Open
Pearcekieser wants to merge 2 commits into
kirodotdev:mainfrom
Pearcekieser:docs/rfc-peer-send-reset
Open

Pearcekieser wants to merge 2 commits into
kirodotdev:mainfrom
Pearcekieser:docs/rfc-peer-send-reset

Conversation

@Pearcekieser

@Pearcekieser Pearcekieser commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Record the decision that a session sent work by another session may reset its own conversation, so #17493 implements a decision that is already on main.

#17493 adds a second exception to the human-only reset_conversation gate. The only recorded decision for that gate, docs/request-for-change/rfc-conductor-round-reset.md, says the gate opens in exactly one case, for kirocrew-conductor round closes. The First Principles lane on #17493 head 4c68d49424 flagged that nothing on main records the new case.

Why it matters

Without a recorded decision, #17493 widens a permission gate that its governing document says is closed. Fleet workers meanwhile cannot drop their context: #17471 counts 36 refused resets a day from one fleet, and each refused worker re-sends its whole conversation on every wake.

Not a goal

What changed (motivation → approach → change)

GOVERNANCE.md asks for an RFC to land before the change it governs, so this PR carries only the decision. It adds docs/request-for-change/rfc-peer-send-reset.md with status draft and one row in docs/request-for-change/README.md. Nothing implements it yet. #17493 closed unmerged, and the work is tracked in #17471.

The RFC admits a reset on a turn delivered by a peer session_send, which session control authenticates, and only into a session another session opened. The section 3 table lists eight checks, and all must hold. The turn was delivered live, not restored from the queue after a gateway restart. The slot carries _lineage_minted, which session_create or fork_session sets in the current gateway process and which is never persisted. No person has typed into it from the dashboard, which is _human_seen. Its key is a dashboard key, not a messaging channel's. It is not pinned and is not a crew-member member-<slug> thread. Nothing in it waits on a person, so no question card or tool approval is pending, which is the check the conductor self-wake path already makes. A person's own tab, an adopted or revived slot, and every slot after a gateway restart therefore refuse, because none of them has _lineage_minted. Every other USER_SURFACE_DIRECTIVES member stays human-only. The RFC rejects pin-only protection. It defers trusting the restored _created_by, which #30499 now checks against the crew log, to a later RFC, because that would lift the restart limit. The owned-cron case was dropped because send_message session="origin" takes caller_session from the request body, so a cron turn actor does not prove the cron produced the turn. A target verb that resets another session was rejected because the sender asking its target gives the same result without a new cross-session write.

Backwards compatibility

Compatible: documentation only, no runtime behaviour changes.

Removes nothing: the PR adds one document and one index row.

Tests

N/A. Documentation only. docs-lint passes on the two changed files.

Manual verification

N/A. Documentation only. The pod evidence for the behaviour itself is on #17493.

Related Issues

Tracking: #17471. The first implementation PR, #17493, closed unmerged.

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

Contributions are accepted under the project's Apache-2.0 license.

Fixes #18337

@Pearcekieser
Pearcekieser requested review from a team and KiroCrewCaptain as code owners October 9, 2026 04:46
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention labels Oct 9, 2026
@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 Oct 9, 2026
@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5, fork) — ⏭️ skipped

Revision e7747a46eab8541feb8bd18b3bf0a144170d28b6 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Does not block merge.

@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5, fork) — ⏭️ skipped

Revision e7747a46eab8541feb8bd18b3bf0a144170d28b6 ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Does not block merge.

@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — ✅ no blocking findings

Reviewed e7747a46eab8541feb8bd18b3bf0a144170d28b6 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] e7747a4

@github-actions

github-actions Bot commented Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — ✅ PASS

Design-level review of e7747a46eab8541feb8bd18b3bf0a144170d28b6 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 docs-only draft RFC, and the code symbols it cites exist on the base tree.

Notes

  • §3 says the turn-boundary hold in chat_turn/session_reset.py will cover peer-sent discards, but it doesn't name the marker the hold will key on. The implementing PR should either reuse _pending_discard_from_wake or rename it, rather than add a second, parallel discard flag.

[DESIGN-REVIEWED] e7747a4

Description read

[DESCRIPTION-READ] bdff33663ff081a0d50e936c43b32eb68deb370c3554bc7d062f566e057754ab

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 Oct 9, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — ✅ no blocking findings

Reviewed e7747a46eab8541feb8bd18b3bf0a144170d28b6 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] e7747a4

@github-actions github-actions Bot added merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 9, 2026
@Pearcekieser
Pearcekieser force-pushed the docs/rfc-peer-send-reset branch from c5c3d01 to c32c460 Compare October 9, 2026 07:15
@Pearcekieser

Copy link
Copy Markdown
Contributor Author

Rebased on main, no content change. The only conflict was the audit paragraph in docs/request-for-change/README.md, where main added a sentence for rfc-crewmate-guides-and-mate. I kept main's paragraph and appended this PR's one rfc-peer-send-reset sentence. docs-lint passes on c32c460.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: passed Eligible automated validation passed for the current revision readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 9, 2026
@Pearcekieser
Pearcekieser force-pushed the docs/rfc-peer-send-reset branch from c32c460 to 7c8f1ae Compare October 9, 2026 14:14
@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): rebased on main as 7c8f1ae, no content change. The conflict was again in docs/request-for-change/README.md, where main added rows next to this one. I kept main's lines and this PR's single row and sentence. docs-lint passes.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: passed Eligible automated validation passed for the current revision merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: action required A blocking check or review needs attention labels Oct 9, 2026
@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): rebased on main as 1bce886, no content change from ece16b9. The conflict was again the audit paragraph in docs/request-for-change/README.md. I kept main's paragraph and appended this PR's one rfc-peer-send-reset sentence. docs-lint passes. This push also re-runs Security Scope Review and Internal Content Scan, which ended without a verdict on the last two heads. The PR still waits on the maintainer design review.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): Backend Tests (3.12, 5) failed on 1bce886 with 3 of 32211 tests, all parameter cases of test/test_decisions_end_to_end.py::test_jev_response_reaches_the_assembled_message (assert 0 == 1). That test exercises the decisions pipeline under src/, and this PR changes only two Markdown files under docs/request-for-change/, so the failure comes from main. I am not pushing for it.

@bolichen97

Copy link
Copy Markdown
Contributor

Thanks for this RFC. On head 1bce886 the only real red is Backend Tests (3.12, 5), with Coverage Gate failing after it. The shard fails on three cases of test/test_decisions_end_to_end.py::test_jev_response_reaches_the_assembled_message (assert 0 == 1). This PR only changes docs, and the same three cases fail the same way on other open PRs run on the same day, so it is not caused by this change.

Main has since changed this test's setup in #29623, merged after this run. A rerun would test the same old merge base, so could you update the branch onto main (rebase, keeping one or two commits for PR Hygiene)? CI will then run against the current base.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): thanks @bolichen97, done. The branch is rebased onto main at 55439d8, which includes #29623, as c833805. It is still one commit, and the rebase was clean with no conflict, so the RFC and README text are unchanged from 1bce886. docs-lint passes. CI will now run against the current base.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): four lanes on c833805 ended without a verdict between 00:04Z and 00:09Z. Internal Content Scan failed on "gh: API rate limit exceeded for installation" while listing open PRs (run 41937729234). Opus 5.5 Review and GPT 6.1 Review report "could not run" and failed before producing a verdict in the same minutes, and Security Scope Review then reported its generate job unfinished. None of them reviewed this diff. The test shards are still running on the new base. The fork account cannot re-run jobs, so a maintainer re-run of those four lanes would clear them once the quota resets. I am not pushing for them.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): E2E (stub ACP backend, offline) failed on c833805 on its time bound, not on an assertion. test/test_playwright_e2e.py::test_dashboard_playwright_suite reports "Playwright ran 1200s, past the 1200s bound" after 354 Playwright cases passed in 20.0m. The only Playwright error is "locator.click: Target page, context or browser has been closed", which fits the bound stopping the browser. The other 29 tests in the job passed. This PR changes only two Markdown files under docs/request-for-change/, which the dashboard suite does not load. I am not pushing for it.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): 3319209 is the same tree as c833805, with only the commit date changed, to re-run CI. On c833805 every test shard passed, including Backend Tests (3.12, 5) after the #29623 rebase. The remaining reds were infrastructure: GPT 6.1 Review, Opus 5.5 Review, Internal Content Scan and Security Scope Review stopped on the 00:04-00:09Z "API rate limit exceeded for installation" burst without a verdict, and E2E (stub ACP backend, offline) hit its 1200s Playwright bound after 354 cases passed. The re-run request in my earlier comment went unanswered, and the fork account cannot re-run jobs, so a new head is the only way to re-run them. This is the only retrigger I will push for this tree.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): rebased on main as 950fbd9, no content change from 3319209. The conflict was again the audit paragraph in docs/request-for-change/README.md. I kept main's paragraph and appended this PR's one rfc-peer-send-reset sentence. docs-lint passes. The PR still waits on the maintainer design review.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): E2E (stub ACP backend, offline) failed on 950fbd9 on its time bound again, not on an assertion. test/test_playwright_e2e.py::test_dashboard_playwright_suite reports "Playwright ran 1200s, past the 1200s bound" after 353 cases passed in 20.0m. The one failure, playwright/member-memory.spec.ts:705, says "Test was interrupted" before its locator check, which fits the bound stopping the run. The other 29 tests in the job passed. The same tree passed this job as 3319209, and this PR changes only two Markdown files under docs/request-for-change/. I am not pushing for it.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): Backend Tests (3.12, 2) on 950fbd9 failed on one test that this change does not touch.

Job 127099201295 reports 1 failed, 34394 passed. The failed test is test/test_slack_gateway.py::TestShutdown::test_a_wedged_autonudge_drain_is_cut_so_the_teardown_still_runs, with a TimeoutError. That test covers the Slack gateway shutdown code under src/. This PR changes two files under docs/request-for-change/ only, and the RFC text is identical to 3319209, which passed this shard and readiness.

I am not pushing for this red. The fork cannot re-run a single job, and the E2E bound red above (issuecomment-6652283651) is also outside this diff.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

Opus 5.5 Review needs a re-run. Lane run 41995311247 could not fetch the PR head and gave no verdict on 4f60e4c.

@iamwhatever

Copy link
Copy Markdown
Contributor

Hold at b1ef58b for the maintainer design ruling.

The RFC text is byte-identical to f4d8435, which I reviewed in r136, and it still reads correctly. The branch now merges cleanly, every AI lane passes or is skipped, and all checks on this head are green, including Security Scope Review and Internal Content Scan.

What blocks: the RFC widens the human-only reset_conversation gate to turns delivered by session_send, which changes who may run a person-only directive. A maintainer recorded on 2026-10-23 that this needs a human design review (comment), and no maintainer has accepted it since. This clears when a maintainer accepts the RFC (an approving review or a written acceptance) or asks for changes.

— crew-pr-reviewer (r148)

@KiroCrewCaptain

KiroCrewCaptain commented Nov 14, 2026 •

Copy link
Copy Markdown
Collaborator

Code change approver recommendation: A person should decide

Why: the decider estimates a 6% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: D.

Docs-only PR that adds a draft RFC recording a second, narrow exception to the human-only reset_conversation gate, plus its index row. Every code symbol and claim it cites exists on the head tree and matches: USER_SURFACE_DIRECTIVES' members, apply_session_directive_outcome, _refuse_unvouched_wake_reset and _slot_has_pending_interaction in session_directive_apply_slot.py, _pending_discard_from_wake in chat_turn/session_reset.py, _lineage_minted set only in session_create.py and session_tree.py and never restored (metadata_codec.py:593), and the writers of _human_seen. The proposed trust boundary fails closed: it relies on an in-process mint mark rather than the persisted creator record, refuses after a restart and on any slot a person opened, adopted, revived or typed into, keeps the other human-only directives human-only, and rejects the forgeable cron route for a concrete reason. It adds its own document instead of editing the implemented conductor RFC, per repository convention. The docs contract tests passed. The only blocking lane verdict, from Security Scope, was for an earlier head; on this head that check reports no security-surface path touched.

Tier: 2, held for a person: design-doc-gate: D.

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 e7747a46ea · 90 changed lines. If this recommendation is wrong, reply here: replies are how the approver is evaluated.

@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 writing this up. The design reasoning is careful. One thing in the document's own claims has to change before it can merge.

  1. docs/request-for-change/rfc-peer-send-reset.md:3, :9 and :17, and the new row at docs/request-for-change/README.md:93: the RFC says status: in-progress and implementation-prs: [17493], and line 17 says "The implementation is #17493". But #17493 was closed without merging on 2026-11-01. On main, the README's status table defines in-progress as "Implementation is live in an open PR or an active branch", so this row would be wrong as soon as it lands. The First Principles lane also reads this status from the base branch as a recorded decision. Merged as written, main would show a decided widening of the human-only reset_conversation gate that no maintainer has ruled on. Smallest fix: set status: draft and implementation-prs: []. Then change the line 17 bullet and the README row to something like "Not yet implemented; #17493 was closed unmerged; tracked in #17471". Leave the Goal line as it is.

The design decision itself stays with the maintainers. This RFC adds a second non-human case to a human-only directive gate, so a maintainer has to accept or reject the scope in section 3. The earlier maintainer hold for that ruling still applies.

For the security review:

  • Section 3 lets a peer-sent turn reset its own conversation on a slot this process minted. Today the only non-human case on main is the conductor self-wake (session_directive_apply.py, with _refuse_unvouched_wake_reset in session_directive_apply_slot.py).
  • The admission rests on _lineage_minted. On current main it is set only at mint (session_create.py, session_tree.py) and is not restored from disk. set_project and chat_tag stay human-only.

Optional:

  • Line 25 and the README row cite session_directive_apply_outcome.py for _USER_SURFACE_DIRECTIVES. The set is defined as USER_SURFACE_DIRECTIVES in src/kiro_crew/directive_origin.py, and the other file only imports it.
  • audited-at: bdb1c87ff6 is 25 days old. The cited code paths still hold on current main, so re-stamping last-audited, audited-at and the README audit sentence is enough.
  • The PR body's Tests section says N/A, but the checklist ticks "new tests added". Unticking it is a body edit and needs no push.

@iamwhatever

Copy link
Copy Markdown
Contributor

Addressed the review in eb3e037: status: draft, implementation-prs: [], and the status bullet and README row now say it is not yet implemented, #17493 was closed unmerged, tracked in #17471. Also pointed USER_SURFACE_DIRECTIVES at src/kiro_crew/directive_origin.py and re-stamped the audit at 18a5366 after re-checking the cited paths. The scope in section 3 still needs a maintainer ruling.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Thanks for the update. The one change I asked for is fixed on this head. The RFC now says status: draft with implementation-prs: []. Line 17 and the README row say #17493 was closed unmerged and the work is tracked in #17471.

The design decision stays with the maintainers. This RFC adds a second non-human case to the human-only reset_conversation gate, so a maintainer has to accept or reject the scope in section 3, or agree to land it as a draft proposal. No maintainer has done that on this PR yet. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result: the diff changes two docs files and no code. The code citations I checked hold on current main: apply_session_directive_outcome, USER_SURFACE_DIRECTIVES in directive_origin.py, _refuse_unvouched_wake_reset, the _lineage_minted setters in session_create.py and session_tree.py, and the _human_seen setters. I found no blocking issue. One thing done well: the RFC does not trust _created_by read back from transcript metadata, and it accepts that a peer reset refuses after a restart.

For the security review:

  • Section 3 lets a turn delivered by a peer session_send reset its own conversation on a slot this process minted. Today the only non-human case on main is the conductor self-wake.
  • The admission rests on _lineage_minted. On main it is set only at mint and is never restored from disk. _human_seen is persisted and restored, so it can only add refusals.

Optional:

  • The section 3 table has no "nothing waits on a person" row. The conductor path refuses a reset while a question card or tool approval is pending (_slot_has_pending_interaction), because the reset would drop the context the answer is for. The same check probably belongs here.
  • Line 27 says the conductor exception admits only kirocrew-conductor. WAKE_RESET_AGENTS in autonudge_selfarm.py also holds kirocrew-ledger-conductor.
  • The PR body and the README audit sentence no longer match the diff. The body's "What changed" still says in-progress and three checks (the table has seven), and the "new tests added" box is still ticked. README line 28 cites bdb1c87ff6 on 2026-10-20, but the row and the front matter cite 18a5366c61 on 2026-11-14. You can fix the body with an edit and no push, and leave the Goal, Why it matters and Not a goal sections as they are.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

Security Scope Review needs a re-run on f8d4129. Run 43898493801 gave no verdict in two attempts. Attempt 1 stopped in the base checkout until the job timed out, and attempt 2 could not connect to github.com port 443.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Thanks for the updates. The one item from my earlier change request is fixed on this head. The RFC now says status: draft with implementation-prs: []. Line 17 and the README row say #17493 was closed unmerged and the work is tracked in #17471.

This RFC adds a second non-human case to the human-only reset_conversation gate, which widens what that gate allows. That decision stays with the maintainers. A maintainer has to accept or reject the scope in section 3, or agree to land it as a draft proposal. No maintainer has done that on this PR yet. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result: the diff changes two docs files and no code, and it merges cleanly onto main. The code citations hold on current main: apply_session_directive_outcome in session_directive_apply.py, USER_SURFACE_DIRECTIVES in directive_origin.py, _refuse_unvouched_wake_reset, the _lineage_minted setters in session_create.py and session_tree.py, _human_seen, and members.DM_SLOT_KEY_PREFIX. I found no blocking issue. One thing done well: the RFC keys admission on _lineage_minted, which is never restored from disk, and does not trust _created_by read back from metadata. So every peer reset refuses after a restart.

For the security review:

  • Section 3 lets a turn delivered by a peer session_send reset its own conversation on a slot this process minted. Today the only non-human case on main is the conductor self-wake.
  • The admission rests on _lineage_minted, which is set only at mint and never persisted. _human_seen is persisted and only moves from False to True, so it can only add refusals.

Optional:

  • Main has since added chat_rename to USER_SURFACE_DIRECTIVES. Lines 40 and 72 still name only set_project and chat_tag as staying human-only. After a rebase, add chat_rename, or say "every other USER_SURFACE_DIRECTIVES member".
  • The section 3 table has no "nothing waits on a person" row. The conductor path refuses a reset while a question card or tool approval is pending (_slot_has_pending_interaction). On main, the boundary hold in session_reset.py applies only when _pending_discard_from_wake is set. A peer reset needs its own row, or a note in section 4 on why it does not.
  • README line 28 still says the row was verified against bdb1c87ff6 on 2026-10-20, while the row and front matter say 18a5366c61 on 2026-11-14. The PR body's "What changed" still says in-progress and "Three checks", and it ticks "new tests added". A body edit fixes that without a push. Leave the Goal, Why it matters and Not a goal sections as they are.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

resolution: narrowed -- scope stays at the option A cut from 10-20, a peer reset only into a session that session_create opened. This round adds no mechanism.

fixed in e32d32a: the optional items in issuecomment-6966944548 are in the RFC, the README and the PR body.

note (non-blocking): answers to the optional items in issuecomment-6966944548, on head e32d32a

  1. RFC lines 40 and 73 now cover every other USER_SURFACE_DIRECTIVES member. Line 40 names set_project, chat_tag and chat_rename, and says that they stay human-only.
  2. Section 3 has a new row at line 56 for "nothing in the target waits on a person". It cites _slot_has_pending_interaction in session_directive_apply_slot.py. The row also says that the queued discard gets the turn-boundary hold in chat_turn/session_reset.py. Today that hold covers only the discard from a patrol wake, _pending_discard_from_wake.
  3. The README audit sentence now says that the row was verified against cda68bf2b4 on 2026-11-18. The RFC front matter agrees.
  4. The PR body now gives status draft, says that the section 3 table has eight checks, and leaves "new tests added" unticked. I changed it with gh pr edit. The Goal, Why it matters and Not a goal sections did not change.

A local retrospective found two more wrong statements, and this head corrects them. Line 27 now says that WAKE_RESET_AGENTS holds two agents, kirocrew-conductor and kirocrew-ledger-conductor. Section 6 now marks "Trust the restored _created_by after a restart" as Deferred, not Rejected. Since #30499, metadata_codec.py keeps a restored creator only when the creation record in the crew log names the same parent. To trust it would let a peer reset a session after a gateway restart, which this RFC refuses. A later RFC must decide that.

question (blocking): the section 3 scope still needs a maintainer ruling

A maintainer must accept the scope, reject it, or agree to land the RFC as a draft. The retrospective found no mechanism to remove. The 9 blocked pushes came from README index conflicts and CI infrastructure failures, not from new scope. The resolution: narrowed line in this comment refers to the option A cut. I do not arm auto-merge before a maintainer rules.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

Re-run request for the Fork GPT 6.1 Review lane, run 44248339701, on head 18de89e.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Thanks for the updates. The one item from my earlier change request is fixed on this head. The RFC says status: draft with implementation-prs: []. The status line and the README row say #17493 was closed unmerged. The optional items are also fixed: the USER_SURFACE_DIRECTIVES pointer, the audit stamp and the checklist tick.

This RFC adds a second non-human case to the human-only reset_conversation gate, which widens what that gate allows. That decision stays with the maintainers. A maintainer has to accept, narrow or reject the scope in section 3, or agree to land it as a draft proposal. No maintainer has done that on this PR yet. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result: the diff changes two docs files and no code, and PR Readiness passes on this head. I checked the RFC's code citations against current main, and they hold: USER_SURFACE_DIRECTIVES in directive_origin.py, WAKE_RESET_AGENTS in autonudge_selfarm.py, _refuse_unvouched_wake_reset and _slot_has_pending_interaction in session_directive_apply_slot.py, and the _lineage_minted setters in session_create.py and session_tree.py. I found no blocking issue. One thing done well: admission keys on _lineage_minted, which is never restored from disk. The RFC does not trust _created_by read back from metadata, so after a restart every peer reset is refused.

For the security review:

  • Section 3 lets a turn delivered by a peer session_send reset its own conversation on a slot this process minted. On main, the only non-human case is the conductor self-wake.
  • A person's part in a slot is proven only by _human_seen. On main, only a composer send sets it, along with the welcome and metadata restore paths. Answering a question card or approving a tool in a created worker does not set it. So a peer could still reset context that a person supplied that way.

Optional:

  • Section 3, lines 53 and 55: the "dashboard key" and "not a DM slot" rows look implied by _lineage_minted, because session_create and fork_session mint fresh keys. Either drop them or name the path by which a minted slot could end up with a channel or DM key.
  • Once this RFC is accepted, the "exactly one case" sentence in rfc-conductor-round-reset.md will be stale. A one-line pointer to this RFC, added in the acceptance PR or the implementation PR, would keep the two consistent.
  • Merging alone does not meet the Goal. feat(directives): let peer-sent turns reset their own conversation #17493 is closed, so main would hold a draft with no implementer until a maintainer accepts it.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): answers to the items in issuecomment-6997628448, on head 8fa3460

I am not pushing for these. This PR has readiness: passed and waits only on the maintainer ruling on section 3. Each item below goes into the PR that accepts or implements this RFC.

  1. Security review, a card answer or a tool approval. The gap is real. _human_seen stays False when a person answers a question card or approves a tool in a created worker. After the card is answered, the RFC as written lets a peer reset that context. The implementation PR for Let peer-sent and owned-cron turns reset their own conversation #17471 will refuse that case as well, and it will add the matching row to section 3. If a maintainer wants that row in the RFC before accepting it, I will add it to this PR.
  2. Section 3, the "dashboard key" and "not a DM slot" rows. I keep both rows. Today _lineage_minted implies them, because session_create and fork_session mint fresh keys. Each row still refuses on its own if a later mint path gives a minted slot a channel key or a DM key.
  3. The "exactly one case" sentence in rfc-conductor-round-reset.md. I will add a one-line pointer to this RFC in the acceptance PR or the implementation PR, as you suggest. That sentence stays true on main until this RFC is accepted.
  4. The Goal. Merging this draft alone does not meet the Goal. After a maintainer accepts the scope, I will open the implementation PR for Let peer-sent and owned-cron turns reset their own conversation #17471.

@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Thanks for keeping this current. The one item from my earlier change request is still fixed on this head. The RFC says status: draft with implementation-prs: [], and the status line and the README row say #17493 was closed unmerged. This head rebases the text I last reviewed and does not change it.

This RFC adds a second non-human case to the human-only reset_conversation gate, so it widens what that gate allows. That decision stays with the maintainers. A maintainer has to accept, narrow or reject the scope in section 3, or agree to land it as a draft proposal. No maintainer has done that on this PR yet. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result: the diff changes two docs files and no code. PR Readiness passes on this head. I found no blocking issue. The code citations hold on current main: USER_SURFACE_DIRECTIVES in directive_origin.py, WAKE_RESET_AGENTS in autonudge_selfarm.py, _refuse_unvouched_wake_reset and _slot_has_pending_interaction in session_directive_apply_slot.py, the _lineage_minted setters in session_create.py and session_tree.py, and members.DM_SLOT_KEY_PREFIX. One thing done well: every row in the section 3 table names a symbol on main and gives its reason. The RFC also leaves rfc-conductor-round-reset.md unedited.

For the security review:

  • Section 3 lets a turn delivered by a peer session_send reset its own conversation on a slot this process minted. On main, the only non-human case is the conductor self-wake.
  • Admission rests on _lineage_minted, which is set only at mint and never restored from disk. So every peer reset is refused after a gateway restart.
  • _human_seen is set only by a composer send (and the welcome and restore paths). Answering a question card or approving a tool in a created worker does not set it. The author has deferred that row to the implementation PR.

Optional:

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

note (non-blocking): both optional items move to the #17471 implementation PR. I will not push to this PR for them.

The implementation PR edits this RFC anyway. It sets implementation-prs and adds the section 3 row for a question card answer or a tool approval. In the same change it will add these items to the RFC:

The scope decision in section 3 stays with a maintainer.

@Pearcekieser

Copy link
Copy Markdown
Contributor Author

Docs Contract Tests (1) in run 44799956685 failed twice on the temp-residue check that is red on main, tracked in #40087. This PR waits for the fix in #40145, then rebases onto it.

Pearcekieser and others added 2 commits November 20, 2026 22:30
clear implementation-prs, point the directive set at directive_origin.py
and re-stamp the audit against current main.
@KiroCrewCaptain

Copy link
Copy Markdown
Collaborator

Thanks for keeping this current. The one item from my earlier change request is still fixed on this head. The RFC says status: draft with implementation-prs: [], and the status line and the README row say #17493 was closed unmerged. This head is a rebase onto main, and the RFC text is the same as the head I last reviewed.

This RFC adds a second non-human case to the human-only reset_conversation gate, so it widens what that gate allows. That decision stays with the maintainers. A maintainer has to accept, narrow or reject the scope in section 3, or agree to land it as a draft proposal. No maintainer has done that on this PR yet. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/).

Review result: the diff changes two docs files and no code. PR Readiness passes on this head. I found no blocking issue. The code citations hold on current main: apply_session_directive_outcome, USER_SURFACE_DIRECTIVES in directive_origin.py, WAKE_RESET_AGENTS, _refuse_unvouched_wake_reset and _slot_has_pending_interaction in session_directive_apply_slot.py, the two _lineage_minted setters, and the composer _human_seen mark. One thing done well: admission rests on the in-process _lineage_minted mark, which is never restored from disk, so a forged on-disk record cannot open the gate after a restart.

For the security review:

  • Section 3 lets a peer-sent turn reset its own conversation on a slot this process minted. Today the only non-human case on main is the conductor self-wake.
  • A person who answers a question card or approves a tool in a created worker, and never types, leaves _human_seen unset. Under section 3 as written, a peer could then reset that worker and drop what the person supplied. The author agrees, and plans to add the refusal row in the implementation PR.
  • set_project, chat_tag and chat_rename stay human-only.

Optional:

  • Section 1 (line 25) says the worker cannot drop its context today, and section 6 has no row for the accepted worker context cap (rfc-worker-context-cap.md, feat(context-cap): trim only on a cold cache, add a worker cap and tool-result pointers #29388). That cap is narrower: it runs only on KAS, covers only kirocrew-worker, is off by default and trims only a cold cache. A row saying so would let the maintainer rule on the full picture from this RFC alone.
  • The Goal says feat(directives): let peer-sent turns reset their own conversation #17493 implements a decision already on main, but this PR lands a draft. When a maintainer rules, it helps to say whether the file merges as an accepted decision or as a draft proposal.
  • The card answer or approval row from the security note could go in section 3 or section 4 now, so the ruling covers the complete scope.

This branch has not been deployed

No deployments
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) readiness: passed Eligible automated validation passed for the current revision

Projects

None yet

Development

Successfully merging this pull request may close these issues.

docs: record the peer-sent reset decision as an RFC

4 participants