principle-subagent-inherits-scope: a brief carries decisions, not just facts and a question

[EdbertChan]

Summary

When handing work to a helper, the guidance names two things to pass on: the facts, and the question.

A decision the person already made is neither, so it gets written as a question again. The rule against leading the helper then rewards re-opening it, because that looks like fairness.

In one measured case the requirement was passed on word for word and still got re-litigated; the helper ranked it fourth of six and argued the premise away.

This adds a third thing to pass on: decisions, stated as constraints.

Review Claim

Guidance for handing off work names decisions as their own category, not to be re-opened.

Review Lane

behavior

Review Unit

corpus-lesson

Safety Invariant

The rule against leading a helper still governs findings; it is narrowed so it no longer applies to choices the person has already made. Nothing becomes un-askable that was open.

Slice Rationale

Prose plus its fires and stays-silent examples. No code changes, so it carries no runtime risk and can land independently of the rest of this stack.

Non-goals

Does not add a check that inspects handoff text before it is sent. The obvious mechanical version was tested and does not work: the requirement was present word for word in the case that went wrong.

Test Plan

Test Plan

Run at this commit:

python3 tests/test_skill_scenarios.py
Ran 6 tests in 0.067s
OK

The two example files are the fires and stays-silent pair for the guidance. No detector is claimed here, because the measurable one was shown not to fire on the real case.

Revert Plan

Revert Plan

Revert this commit. The guidance returns to two categories. No behaviour changes either way.

---

Note

Low Risk
Corpus guidance and example prose only; no executable behavior or enforcement logic changes.

Overview
Extends principle-subagent-inherits-scope so parent handoffs must pass three items—facts, the question, and settled user decisions—not just facts plus a question.

The skill now tells parents to phrase decisions as explicit constraints (not open questions), clarifies that anti-priming applies to findings only (omitting a decision is not neutral), and adds requirements-traceability prior art. fires_example.md adds a third drift case where verbatim requirements still get re-litigated when filed under the question, with an explicit note that mechanical containment checks do not catch it. stays_silent_example.md adds the matching good handoff where the decision travels as a non-reopenable constraint.

Documentation and scenario examples only; no runtime or detector changes.

^{Reviewed by Cursor Bugbot for commit 5fe7b3c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[EdbertChan]

This pull request is part of a Mergify stack:

#	Pull Request	Link
1	wrong-check-reflect: one shot per reply, and scope the reflect scan to the turn	#796
2	reflect: catch an evidence-order correction from the ledger transition, not from wording	#782
3	unverified-tag-ledger: gate the next-prompt reminder, default to stale only	#783
4	diu-stop: a file citation buys silence only when it is backed	#784
5	diu-stop: the marker gate reads prose, so a mention is not a use	#785
6	cat-mode: show the escape-hatch tag in the form the gate accepts	#786
7	principle-subagent-inherits-scope: a brief carries decisions, not just facts and a question	#788	👈
8	llm-judge: claude is the only default runner	#793

[EdbertChan]

Revision history

#	Type	Changes	Reason	Date
1	initial	83d83f9		2026-09-22 18:38 UTC
2	rebase	83d83f9 → 843cbad (rebase only)		2026-09-22 18:38 UTC
3	rebase	843cbad → 350ee29 (rebase only)		2026-09-23 14:06 UTC

[EdbertChan]

Mergify repair stopped: lowest open stack PR #788 is based on stack/EdbertChan/reflect/subagent-decisions-slot-20260922/cat-mode-show-escape-hatch-tag-form-gate-accepts--f47021e8, which still belongs to open PR(s) #786 outside this stack; leaving it alone to avoid dropping dependency changes

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_7e2c1b74-2ce0-46c4-9194-fd92803fec92)

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_57e573f5-63a4-489d-8b8a-def8e0084b09)

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_22c5d3f9-d56a-4781-aec3-4e1d9c772d57)

[mergify]

Queued — the merge queue status continues in this comment ↓.

[EdbertChan]

@Mergifyio queue

[mergify]

Merge Queue Status

🛑 Queue command has been cancelled

[EdbertChan]

@Mergifyio queue

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_3a54a118-c79c-4eeb-a702-06ade81d917e)

[mergify]

queue

☑️ Command queue ignored because it is already running from a previous command.

[EdbertChan]

@Mergifyio queue

[mergify]

Merge Queue Status

• ✅ Entered queue — 2026-09-24 12:26 UTC · Rule: admin-bypass · triggered by @EdbertChan with the @mergifyio queue command
• ✅ Checks passed · on draft merge queue: checking #788 on main (58acd2b) #948
• ✅ Merged — 2026-09-24 12:44 UTC · at cf30493e534de6a0266779845f7c224784a1cdab · squash

This pull request spent 18 minutes 27 seconds in the queue, including 17 minutes 43 seconds running CI.

Required conditions to merge

• check-success = lint
• check-success = test
• check-success = validate