Skip to content

docs(rfc): conductor default patrol on a crew- or member-mode conductor - #18699

Closed
rubencu wants to merge 1 commit into
kirodotdev:mainfrom
rubencu:docs/rfc-conductor-default-patrol-crew-member
Closed

rubencu wants to merge 1 commit into
kirodotdev:mainfrom
rubencu:docs/rfc-conductor-default-patrol-crew-member

Conversation

@rubencu

@rubencu rubencu commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: record, in the RFC that decided otherwise, the decision to arm the gateway's default patrol on a crew- or member-mode conductor, so the implementation (#18592) can be judged against a decision on the base.

docs/request-for-change/rfc-conductor-default-patrol.md (status in-progress) lists "Arming on a crew- or member-mode conductor" as a non-goal and makes the refusal an exit criterion, because the bind route cannot prove the turn is the session's own. Two days after it landed, the patrol's owner filed #17603 asking for the patrol on exactly those conductors. Member DM threads are the normal conductor shape, so the safety net the patrol is has been inert where conductors run: on a live gateway every refusal since 2026-09-24 was a member-mode conductor's bind (3 refusals, 0 arms).

The implementation is open as #18592. Its First Principles lane reads the RFC from the base branch and blocks on this exact point (product shape without a recorded decision), and its Design lane asks that the RFC be amended in a PR that lands before it with the patrol owner's sign-off. GOVERNANCE.md says the document lands first and the implementation references it. This PR is that document, proposed for the patrol's owner to accept by merging.

Why it matters

Member DM threads are the normal conductor shape. Until this decision is on the base, every member-mode conductor's workers report to a ledger nobody reads, and the fix for that cannot merge: its First Principles lane blocks on the missing decision by design.

Not a goal

What changed

  • docs/request-for-change/rfc-conductor-default-patrol.md
    • Status: a revision line for 2026-10-10 naming Member/crew conductors get no default patrol on bind #17603 as the ask and fix(conductor-patrol): admit the gateway patrol on a crew/member slot #18592 as the implementation.
    • §2: the crew/member conductor moves from the non-goals to the goals, as the gateway's own content-pinned arm; the non-goal becomes "giving the bind route turn provenance, or self-arming a crew/member conductor from it".
    • §3: a paragraph on how such a slot admits the patrol: the flag only with the fixed text and the slot's own work-ledger watch at arm time, a gateway-patrol trust entry beside the row's tag, the content re-read against the stored row at fire time (the loop store is agent-writable and the entry names only an id and a slot), monitor_update tuning bounds and banner but not message or watch, and the two trust-entry kinds never vouching for each other so the patrol's wake cannot reset_conversation.
    • §4: the agent-writable-tag risk now covers the crew/member boundary (the tag is a hint there, not authorization), plus the patrol the conductor cannot reword.
    • §5: the default arm is the gateway's arm, what a cron-injected or sub-agent-shared bind can start on such a slot (the fixed patrol on the slot's own ledger, nothing else), and what the audit records.
    • §6: the self-arm route Member/crew conductors get no default patrol on bind #17603 suggested (directive consumer or runner turn end) and the first decision's refusal are recorded as alternatives with the reasons they were not taken.
    • §8: fix(work-ledger): arm a work-ledger patrol on a conductor's bind when it has none #17069's criteria keep the first list minus the refusal; fix(conductor-patrol): admit the gateway patrol on a crew/member slot #18592's criteria are listed, each pinned by a test in that PR.
    • Front matter: implementation-prs adds 18592, tracking-issues adds 17603, re-audited at 950be41684.
  • docs/request-for-change/README.md: the row is re-audited against 950be41684 (the bind-time arm from fix(work-ledger): arm a work-ledger patrol on a conductor's bind when it has none #17069 is on main; the crew/member admission is not: no is_gateway_patrol, no gateway_patrol entry kind), and the audit-history paragraph gains the matching sentence.

If the owner prefers the self-arm route the issue suggested, this document and #18592 change together; the decision is theirs, and merging this records it.

Tests

scripts/docs-lint.sh clean; test/test_pr_quality_gates.py passes (76). Docs only.

Related Issues

Refs #17603. Implementation: #18592.

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

@rubencu
rubencu requested review from a team and KiroCrewCaptain as code owners October 10, 2026 03:04
@rubencu
rubencu requested a review from chenmingwei23 October 10, 2026 03:04
@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 10, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from 9f6144f to 00013b1 Compare October 10, 2026 03:35
@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 10, 2026
@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5, fork) — ⏭️ skipped

Revision 394624e639f74b9a83449aa00332cec4a56a7ea3 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

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

Revision 394624e639f74b9a83449aa00332cec4a56a7ea3 ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Advisory — does not block merge.

@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from 00013b1 to e1c2904 Compare October 10, 2026 03:56
@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — 🟡 CONCERNS

Design-level review of 394624e639f74b9a83449aa00332cec4a56a7ea3 via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

The RFC stores gateway_patrol trust entries in the same file as self-arm entries, and today's reader of that file ignores the entry's kind, so a rollback turns patrol entries into self-arm vouches.

Watch

  • The two entry kinds share one record, and the existing reader can't tell them apart. §3 says the entry goes "in the same keystone-gated record the self-arm uses (autonudge_selfarm, entry kind: gateway_patrol)" and that "the two entry kinds never vouch for each other."
    • On the base, the record is {loop_id: {"slot_key", "armed_ts"}} and has no kind field. is_recorded_self_arm (autonudge_selfarm.py:158) matches on loop id and slot only.
    • That separation only holds while fix(conductor-patrol): admit the gateway patrol on a crew/member slot #18592's kind-aware reader is deployed. A revert or downgrade leaves the gateway_patrol entries on disk, and the old reader treats each one as a self-arm.
    • The loop store is agent-writable. An agent can set self_armed: true and rewrite the message on the patrol row. The old fire guard then admits that outside text into a member's thread, and the wake-reset gate opens.
    • The RFC is where this persisted-format choice is made, and it has no story for rollback or for legacy entries that carry no kind.
    • Fix: put gateway-patrol entries in a separate record file that the self-arm reader never opens. Otherwise, state that an entry with no kind means self-arm, and say what a rollback does with gateway_patrol entries.
    • Clears when: §3/§5 name a separate record file for gateway-patrol entries, or record the legacy-entry and rollback rule and pin it in #18592's exit criteria.

[DESIGN-REVIEWED] 394624e

Description read

[DESCRIPTION-READ] 4a1de34dd47b74f76ba0c6e851248c50ac44295bf016f5fffe84b1104c469c1f

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 added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 10, 2026
@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — ✅ no blocking findings

Reviewed 394624e639f74b9a83449aa00332cec4a56a7ea3 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 394624e

@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — ✅ no blocking findings

Reviewed 394624e639f74b9a83449aa00332cec4a56a7ea3 via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 394624e

@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 10, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from e1c2904 to 650d35a Compare October 10, 2026 07:34
@github-actions github-actions Bot added merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed merge conflict Branch has merge conflicts with its base — author must resolve before merge labels Oct 10, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from 650d35a to 156808e Compare October 11, 2026 06:01
@rubencu

rubencu commented Oct 11, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main at 2e2bd0679f (one commit, 156808ef83566d74e8f9cf252daf21fc08cdeaaf, lease 650d35aab4). The conflict was the shared audit paragraph in docs/request-for-change/README.md: main's paragraph (now ending with the rfc-wake-on-stop-signal sentence) is kept whole and this PR's rfc-conductor-default-patrol sentence follows it; this PR's table row is unchanged and rfc-conductor-default-patrol.md is byte-identical to 650d35aab4. Docs lint, the fact-check and PR quality-gate tests (174), the comment-history gate and git diff --check pass.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision 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 Oct 11, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from 156808e to 6f1e51f Compare October 11, 2026 07:18
@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 merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 11, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from eacd05d to 0f8dc87 Compare October 12, 2026 10:06
@rubencu

rubencu commented Oct 12, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main at e5a0b3d3b2 (eacd05dda6 -> 0f8dc87804), same single commit: #19287 extended the README audit paragraph again, so this change's re-audit sentence is re-appended after main's text. rfc-conductor-default-patrol.md is byte-identical to the previous head (status: in-progress). docs-lint and git diff --check pass.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision 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 Oct 12, 2026
@rubencu
rubencu force-pushed the docs/rfc-conductor-default-patrol-crew-member branch from 0f8dc87 to 336088f Compare October 12, 2026 16:26
@rubencu

rubencu commented Oct 12, 2026

Copy link
Copy Markdown
Contributor Author

Rebased onto main at 0cda1525d8 (0f8dc87804 -> 336088fefb), same single commit: #19352 and #19374 extended the README audit paragraph, so this change's re-audit sentence is re-appended after main's text. rfc-conductor-default-patrol.md is byte-identical to the previous head (status: in-progress). docs-lint and git diff --check pass.

@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 labels Oct 12, 2026
@rubencu

rubencu commented Oct 12, 2026

Copy link
Copy Markdown
Contributor Author

Rebased 336088fefb -> 53a6f29226 on main 63c155a582 after #19043 extended the README audit paragraph and added the rfc-crewmate-dashboard-upkeep-skill row. Resolution as before: main's paragraph whole with this RFC's re-audit sentence appended, both table rows kept, rfc-conductor-default-patrol.md byte-identical with status: in-progress. docs-lint, git diff --check and the 1,400 tests that read docs/request-for-change pass. MERGEABLE.

@rubencu

rubencu commented Oct 13, 2026

Copy link
Copy Markdown
Contributor Author

Rebased 53a6f29226 -> 83a788b844 on main ec16b839bd after #18681 extended the README audit paragraph. Same resolution: main's paragraph whole with this RFC's re-audit sentence appended, rows kept, rfc-conductor-default-patrol.md byte-identical with status: in-progress. docs-lint, git diff --check and the 1,420 tests that read docs/request-for-change pass. MERGEABLE.

@rubencu

rubencu commented Oct 13, 2026

Copy link
Copy Markdown
Contributor Author

Rebased 83a788b844 -> 5d1beabbed after #19488 and #19609 extended the README audit paragraph; same resolution, rfc-conductor-default-patrol.md byte-identical with status: in-progress, docs-lint and the 1,420 README-reading tests pass. MERGEABLE.

@iamwhatever

Copy link
Copy Markdown
Contributor

This docs-only PR revises rfc-conductor-default-patrol.md so a crew- or member-mode conductor also gets the bind-time patrol, as the gateway's own content-pinned arm with a gateway-patrol trust entry that is checked again at fire time. The patrol's wake never becomes a self-arm and cannot reset_conversation. Head 5d1beab is a rebase, and the RFC file is byte-identical to the one I reviewed before. The README row now correctly says that on main a crew/member bind commits with 200 and patrol: "refused" plus patrol_note, so my earlier wording request is met. The security reasoning still holds: only fixed gateway text on the slot's own ledger can reach a member thread this way, and a forged tag without an entry or a rewritten message, banner or watch is refused at fire time. The merge still turns on the question to the patrol's owner, iamwhatever, which nobody has answered yet. Do you accept the gateway-arm route in place of the self-arm through the session-directive consumer or the runner's turn end that you asked for in #17603? Second, the current head's Design CONCERNS has no reply. §4 accepts that editing the patrol text leaves stored crew/member patrols refused at every fire while still counting as active, so unpatrolled stops warning, which is the silent failure this RFC exists to prevent. Please either have the RFC treat a content-stale default patrol as absent, both for ensure_patrol re-arming on the next bind and for the unpatrolled flag, and add that as a #18592 exit criterion, or post a disposition explaining why the digest-test guard is enough. — crew-pr-reviewer (deep re-check)

The first version of rfc-conductor-default-patrol.md listed arming on a
crew- or member-mode conductor as a non-goal, because the bind route
cannot prove the turn is the session's own. Two days after it landed the
patrol's owner filed kirodotdev#17603 asking for the patrol on exactly those
conductors, which are the normal conductor shape; on a live gateway
every refusal since 2026-09-24 was a member-mode conductor's bind.

This revision records the decision kirodotdev#17603 asks for and how it is kept
inside the property section 5 protects. The patrol is admitted on such a
slot as the gateway's own arm, never a self-arm: pinned on its fixed
text and the slot's own work-ledger watch at arm time, backed by a
gateway-patrol trust entry, and re-read against the stored row at fire
time, so a message rewritten in the agent-writable store is refused like
any other outside text. The self-arm route the issue suggested is
recorded as an alternative with the reasons it was not taken. Sections
2, 3, 4, 5, 6 and 8 change; the README row is re-audited against
950be41. The implementation is kirodotdev#18592, which rebases onto this
document so its First Principles lane reads the decision from the base.

Refs kirodotdev#17603
@rubencu

rubencu commented Oct 13, 2026

Copy link
Copy Markdown
Contributor Author

Rebased 5d1beabbed -> 394624e639 after the RFC README audit paragraph moved again; main's paragraph is kept whole with this RFC's re-audit sentence appended. rfc-conductor-default-patrol.md is byte-identical with status: in-progress; docs-lint, git diff --check, 1,423 README-reading tests and push_guard pass. MERGEABLE.

@iamwhatever

Copy link
Copy Markdown
Contributor

This docs-only PR revises rfc-conductor-default-patrol.md so a crew- or member-mode conductor also gets the bind-time patrol, as the gateway's own content-pinned arm with a gateway-patrol trust entry that is checked again at fire time. Head 394624e is another rebase: the RFC file is byte-identical to 5d1beab, and the README row still correctly reports the 200 + patrol: "refused" behaviour on main. The Design Review on this head raises a new CONCERNS, with no ai-review-disposition reply yet. I checked it against main and it holds. §3 puts the kind: gateway_patrol entry in the same autonudge_selfarm record as self-arm entries. But is_recorded_self_arm (src/kiro_crew/autonudge_selfarm.py:158-166 on main) matches only loop id and slot and never reads kind. So after a revert or downgrade of #18592, every patrol entry vouches as a self-arm. And because the loop store is agent-writable, setting self_armed: true and rewriting the patrol row's message would get outside text into a member's thread and open reset_conversation. Please post a disposition, and preferably have §3/§5 put gateway-patrol entries in a separate record file the self-arm reader never opens, or state the legacy/no-kind and rollback rule and pin it in #18592's exit criteria. The earlier question to the patrol's owner, iamwhatever, is still unanswered and still decides the merge: do you accept this gateway-arm route in place of the self-arm through the session-directive consumer or the runner's turn end that you asked for in #17603? — crew-pr-reviewer (deep re-check)

@rubencu

rubencu commented Oct 14, 2026

Copy link
Copy Markdown
Contributor Author

Disposition: the new rollback finding holds. §3 currently shares the self-arm record, while main is_recorded_self_arm matches only loop id and slot and ignores kind; reverting/downgrading #18592 could therefore reinterpret a gateway-patrol row as self-arm authority. I am not choosing the separate-record-file fix versus an explicit legacy/no-kind rollback rule here because #18592 remains sibling-owned for design and the patrol owner’s gateway-arm-versus-self-arm decision is still open. Both choices are now one exact needs-human item; no RFC text changed in this pass.

@rubencu

rubencu commented Oct 14, 2026

Copy link
Copy Markdown
Contributor Author

No rebase pushed: current main now carries the accepted v2 RFC (#19742) whose section 9 says member/crew slots get no outside arm, while this revision admits the gateway patrol on those slots. That is a direct design conflict, not a mechanical README merge. I aborted the local rebase and left fork head 394624e639 unchanged; the gateway-arm route plus rollback-record choice remains the needs-human decision recorded above.

@rubencu

rubencu commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor Author
  • Gateway-patrol route conflicts with the RFC owner’s accepted design and shares a rollback-unsafe self-arm record — accepted; route withdrawn

The author accepts §2 non-goals, §5 and §6 of RFC revision #19893 (2c237a2a75) and the review finding that is_recorded_self_arm ignores kind. The gateway-arm route is abandoned. The member-slot patrol defect will be re-approached as a self-arm provenance change in the session-directive consumer, per #17603. This PR is withdrawn; maintainer, please close it.

@iamwhatever

Copy link
Copy Markdown
Contributor

Thank you @rubencu for this RFC revision and for keeping it rebased so patiently.

Closing as you asked in your comment:

The author accepts §2 non-goals, §5 and §6 of RFC revision #19893 (2c237a2a75) and the review finding that is_recorded_self_arm ignores kind. The gateway-arm route is abandoned. The member-slot patrol defect will be re-approached as a self-arm provenance change in the session-directive consumer, per #17603. This PR is withdrawn; maintainer, please close it.

The RFC revision on main (#19893, commit 2c237a2, with #19742) lists "Arming on a crew- or member-mode conductor" as a non-goal, see docs/request-for-change/rfc-conductor-default-patrol.md:41.

Your branch is kept, and you are welcome to reopen or open a new PR when the self-arm route is ready. Thanks again!

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) 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.

2 participants