Promote Develop: Resync-Filed Model Tiers, Handoff Lifecycle, Hub-Only Tests, and Copilot Transient Fixes by ptr727 · Pull Request #3544 · ptr727/ProjectTemplate

ptr727 · 2026-10-28T00:16:11Z

Summary

Promotes develop to main, carrying round 1 of the resync-issue burndown:

Make the Model Tiers Bullets Agree on What a Dispatch Naming No Model Runs At #3538: AGENTS.md "Model Tiers" now states the no-model dispatch result once, conditionally, in the highest bullet. The two bullets that contradicted it lost their unconditional claims. The three skills including the section are regenerated.
Bring the Durable Knowledge Handoff Bullets Onto the Standing-Issue Lifecycle #3539: GOVERNANCE.md "Durable Knowledge and Self-Improvement" describes the standing-issue handoff lifecycle, matching AGENTS.md "Session Scope". agent-conduct is regenerated.
Keep Hub tests/ Paths Out of the Hub-Only Detector #3540: spec/files.json gains a repoOwnedTrees list declaring tests, and spec/audit.py hub_only_paths leaves those paths out. A carrier's own tests/ file is no longer reported as hub-only. spec/validate.py shape-checks the list, and AUDIT.md, RESYNC.md and the project-types assertion state the exclusion.
Re-Request Once on a Transient Copilot Fetch Error #3542: pr_review.py wait and status read the Copilot run log before calling an error refusal quota. A too many files changed line is transient and earns one re-request, as COPILOT_ERROR_TRANSIENT on the existing exit 41.
Refuse a Repo-Owned Tree Overlapping Carried Content #3545: promotion fix for Copilot's finding on this pull request. spec/validate.py refuses a repoOwnedTrees source that overlaps a carried trees[].source or holds a baseline path.

Fixes #3529
Fixes #3504
Fixes #3528
Fixes #3510

🤖 Generated with Claude Code

… Runs At (#3538) ## Summary `AGENTS.md` "Model Tiers" gave two answers for a dispatch that names no model. The `highest` bullet states the conditional form: the agent definition's model where one is set, otherwise the session's own. Two later bullets stated only the session fallback. - The "A tier resolves only through this table" bullet loses its trailing clause. It now says only that such a session names no model. - The "An orchestrating seat" bullet loses its restating sentence. Its rationale, "a choice nobody made", is false where an agent definition pins a model. The `highest` bullet stays the single statement of the result. `scripts/build_dist.py` regenerated the include regions in `backlog-burndown`, `local-strict-review`, and `unattended-handoff`, plus their generated copies. ## Pre-existing findings, filed - `AGENTS.md` "Model Tiers", the paragraph under the tier table, and the `spec/model-tiers.json` note: an unconditional "session's own model" claim for a tool with no row. Class `pre-existing`, filed as #3537, since the two change together and the spec file is outside this change. Closes on promotion: #3529 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…ifecycle (#3539) ## Summary `GOVERNANCE.md` "Durable Knowledge and Self-Improvement" described the old per-link handoff lifecycle in two bullets. This brings both onto the standing-issue lifecycle that `AGENTS.md` "Session Scope" and the `session-handoff` skill state. - The blocker bullet said a handoff "closes with its successor". It now says the next round replaces the handoff's body in place, which is why a wait belongs in the backlog instead. - The `blocked` bullet said the label comes off "when a session hands the link back to be worked". It now comes off once a session writes the next round, matching `session-handoff`'s resume step 6. - The `agent-conduct` include region and its generated copies are regenerated with `scripts/build_dist.py`. ## Dropped findings - `GOVERNANCE.md:20` (and line 103 of the three `agent-conduct` copies), `style`: five untouched sentences in the blocker bullet exceed 25 words. The diff-mode sentence-length check reads the whole changed line. A later sentence-length round owns them. Closes on promotion: #3504 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

## Summary `spec/audit.py` reported a carrier's own `tests/<name>.py` as `hub-only:` whenever the hub has a test module of the same name. Its remedy then suggested deleting the file. The hub's `tests/` tree is never carried, so such a hit is always a path collision. This closes the class rather than adding one `gaps` entry per instance. - `spec/files.json` gains a `repoOwnedTrees` declaration, listing `tests` with its reason. - `spec/files.schema.json` describes the new array. - `hub_only_paths` drops every hub path below a declared repo-owned tree. Both of its consumers inherit this: the audit's hub-only check and the fidelity-honesty manifest gap pass. - A new `hub_paths_below` helper holds the declared-source walk, shared by both `hub_only_paths` loops and `check_tree_verbatim`. - `AUDIT.md`'s carried-scope bullet, the `carried.hubonly.absent` assertion in `spec/project-types.json`, and `RESYNC.md` section 4 each name the exclusion. - `spec/validate.py` checks the `repoOwnedTrees` shape beside its `trees` checks. It refuses a missing source or reason, a source at or above the repository root or not in normalized form, a source with a symlinked component, and a source that is not a directory. - `tests/test_audit.py` adds a case where a carrier's `tests/test_a.py` shares a constructed hub test's name and is not reported. A control shows a hub-only path outside `tests/` is still reported. Its fixture loads the shipped declaration, so a malformed entry or a dropped `tests` fails the suite. - `tests/test_spec_validate.py` adds `RepoOwnedTreeGateCase`, running the real gate against a scratch tree for each refusal and a plain entry. It shares a new `run_gate_on_files` runner with `TreePathGateCase`. ## Verification - The full unittest suite passes under the OPERATIONS.md invocation. - With the `tests` declaration renamed, both new audit cases fail. - `ruff check`, `ruff format --check`, `mypy`, `spec/validate.py`, and `scripts/prose_lint.py --diff origin/develop` are clean. ## Pre-existing findings, not filed - `AUDIT.md:88`, `pre-existing`: `prose_lint.py --check sentence-length` flags three sentences of 36, 53, and 35 words in the carried-scope bullet. This change added one 25-word sentence there and wrote none of the three. - `RESYNC.md:92`, `pre-existing`: the same check flags a 41-word sentence in section 4. This change added a shorter sentence before it. - `AUDIT.md:88`, `RESYNC.md:92`, `spec/project-types.json:161`, `pre-existing`: each derivation names the baseline and now the repo-owned exclusion, but not the carried `trees` exclusion `hub_only_paths` has always applied. ## Dropped findings - `scripts/canonical_review.py:303`, consolidation on text this change did not touch: its tree walk could reuse `hub_paths_below`. Closes on promotion: #3528 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

## What `pr_review.py` read every Copilot error refusal as a possible quota hit, so `wait` stopped and handed the drive to the maintainer. The same body is posted for a transient fetch failure that one re-request clears. `wait` and `status` now read the reviewer's run log for a `too many files changed` line before taking the quota reading. - `run_cause` also returns the log's transient-failure line, beside the error type and rate-limit sentence it already read. A logged rate limit still wins. - One primitive, `transient_state`, says where an error refusal stands on its one re-request: `owed`, `pending`, `sent`, or `spent`. `wait`'s request, its exit, and the digest line all read it. - `wait` sends the one re-request while the state is `owed`. A request made after the refusal spends it, so a request that vanished is not sent again. Two error refusals in a row spend it too. - A pending re-request polls until the next round lands, then exits on that round. With no round by the timeout it exits `30`, `status=PENDING`. - A held head, into a branch other than the default, exits `41` as `COPILOT_ERROR_TRANSIENT`, naming `--request` as what sends the re-request. Elsewhere an owed refusal steps aside, so an unrecorded request, an outside answer, or the repo-wide quota keeps its own status. - A spent or sent-and-unanswered re-request, an unread log, or a log naming neither cause keeps `COPILOT_ERROR_POSSIBLE_QUOTA`. Its status line now says which reading was taken. - `status` prints a `RUN LOG READING` line under an `ERROR` refusal, naming the reading taken. No exit code is added. The transient state rides the existing `0`, `30`, `40`, and `41` outcomes and their status lines. The help text, `scripts/README.md`, and `pr-review-conduct` describe the transient exception. The skill's generated copies are regenerated. `drive-pr` carries no changed section, so its copies are unchanged. ## Tests `tests/test_pr_review.py` covers the reading with constructed run logs and request histories: the transient line parsed with and without an error type, a rate limit outranking it, one re-request reviewed normally on the current head and on an earlier one, a second refusal ending the poll as quota, two refusals in a row, a pending re-request at the timeout, a re-request already sent, an unrecorded re-request, a second wait on a pending re-request, a held head, an unread log, and the digest's readings. ## Local strict review findings The first push left two `introduced` findings open once its edit budget was spent. Both are fixed in dc2bbe5, and that push's pass raised none. - `scripts/README.md:225`, `introduced`, fixed: it said `wait` exits `41` as `COPILOT_ERROR_TRANSIENT` wherever it sent none, while only a held head reaches that exit. - `scripts/pr_review.py` `transient_state`, `introduced`, fixed: the digest and the exit each read the request history. The read is now kept on the payload, so they share one answer. ## Filed elsewhere - #3541: the digest's fixed `COPILOT REFUSED THIS ROUND` text says "re-requesting repeats it", which a transient refusal contradicts, and the `RUN LOG READING` line belongs inside that block. Both need `scripts/pr_review_digest.py`, outside this change's files. Closes on promotion: #3510 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

coderabbitai · 2026-10-28T00:16:18Z

Important

🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

Source mode: Auto
Repository: ptr727/ProjectTemplate
Branch: develop
File: .coderabbit.yaml
Commit: e1c3a01
Layers: Repository YAML
Review profile: CHILL
Plan: Advanced
Run ID: 580146ec-7d16-4856-87d3-ebdc861b2224

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

codecov · 2026-10-28T00:21:35Z

Codecov Report

❌ Patch coverage is 68.51852% with 34 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.50%. Comparing base (d74cf30) to head (e1c3a01).
⚠️ Report is 432 commits behind head on main.

Files with missing lines	Patch %	Lines
spec/validate.py	0.00%	33 Missing ⚠️
spec/audit.py	88.88%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3544      +/-   ##
==========================================
- Coverage   81.58%   81.50%   -0.09%     
==========================================
  Files          26       26              
  Lines        8972     9065      +93     
==========================================
+ Hits         7320     7388      +68     
- Misses       1652     1677      +25

Flag	Coverage Δ
python-3.13	`81.50% <68.51%> (-0.09%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Copilot

🟡 Changes recommended

A critical active finding remains because repo-owned sources can overlap carried declarations, allowing contradictory paths to suppress hub-only findings without test coverage for the conflict.

1 open issue

Reject repo-owned sources overlapping carried tree declarations · New

What changed in this PR

This pull request promotes develop to main, synchronizing governance guidance, generated skills, repository-owned audit paths, and transient Copilot review handling.

Changes:

Aligns model-tier dispatch and handoff lifecycle guidance across canonical and generated instruction files.
Adds repoOwnedTrees support to manifests, schemas, validation, auditing, documentation, and tests.
Adds Copilot run-log classification and one bounded re-request for transient failures.
Review findings remain for broad transient-error matching, capped review-history detection, and missing ownership-overlap validation.

File	Description
`tests/test_spec_validate.py`	Adds validation coverage for repository-owned tree declarations.
`tests/test_pr_review.py`	Tests transient Copilot failure detection, retry state, quota handling, and status output.
`tests/test_audit.py`	Verifies repository-owned test paths are excluded from hub-only findings.
`spec/validate.py`	Validates `repoOwnedTrees` declarations.
`spec/project-types.json`	Documents the repository-owned tree exclusion.
`spec/files.schema.json`	Defines the `repoOwnedTrees` schema.
`spec/files.json`	Declares the `tests` tree as repository-owned.
`spec/audit.py`	Applies repository-owned exclusions to hub-only auditing.
`scripts/README.md`	Documents transient Copilot refusal handling.
`scripts/pr_review.py`	Reads run logs and manages one transient re-request.
`RESYNC.md`	Documents repository-owned paths in resync guidance.
`GOVERNANCE.md`	Aligns blocker and handoff lifecycle guidance.
`AUDIT.md`	Documents repository-owned tree audit behavior.
`AGENTS.md`	Removes contradictory model-tier dispatch claims.
`.github/skills/unattended-handoff/SKILL.md`	Regenerated model-tier guidance.
`.github/skills/pr-review-conduct/SKILL.md`	Documents transient Copilot refusal handling.
`.github/skills/local-strict-review/SKILL.md`	Regenerated model-tier guidance.
`.github/skills/backlog-burndown/SKILL.md`	Regenerated model-tier guidance.
`.github/skills/agent-conduct/SKILL.md`	Regenerated handoff lifecycle guidance.
`.claude-plugin/fleet-skills/skills/unattended-handoff/SKILL.md`	Regenerated plugin skill distribution.
`.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md`	Regenerated plugin review guidance.
`.claude-plugin/fleet-skills/skills/local-strict-review/SKILL.md`	Regenerated plugin skill distribution.
`.claude-plugin/fleet-skills/skills/backlog-burndown/SKILL.md`	Regenerated plugin skill distribution.
`.claude-plugin/fleet-skills/skills/agent-conduct/SKILL.md`	Regenerated plugin handoff guidance.
`.claude-plugin/fleet-skills/.source-digests/unattended-handoff`	Updates the generated skill digest.
`.claude-plugin/fleet-skills/.source-digests/pr-review-conduct`	Updates the generated skill digest.
`.claude-plugin/fleet-skills/.source-digests/local-strict-review`	Updates the generated skill digest.
`.claude-plugin/fleet-skills/.source-digests/backlog-burndown`	Updates the generated skill digest.
`.claude-plugin/fleet-skills/.source-digests/agent-conduct`	Updates the generated skill digest.
`.agents/skills/unattended-handoff/SKILL.md`	Updates the canonical model-tier guidance.
`.agents/skills/pr-review-conduct/SKILL.md`	Updates the canonical Copilot review guidance.
`.agents/skills/local-strict-review/SKILL.md`	Updates the canonical model-tier guidance.
`.agents/skills/backlog-burndown/SKILL.md`	Updates the canonical model-tier guidance.
`.agents/skills/agent-conduct/SKILL.md`	Updates the canonical handoff lifecycle guidance.

_{Review effort: Lite}

## Summary This is a promotion fix for review thread `PRRT_kwDOQ5caqM6w8BW3` on #3544. `spec/validate.py` now refuses a `repoOwnedTrees` source that overlaps carried content. The audit skips every hub file below a repo-owned source. An overlap would therefore hide hub-only findings while the manifest still claims the content is carried. The check refuses three shapes: - a source equal to a `trees[].source` - a source that is an ancestor or a descendant of a `trees[].source` - a source with a `baseline` path at or below it `RepoOwnedTreeGateCase` runs the real gate on a scratch tree for each refused shape, plus a control beside the carried content. The fixture paths are constructed. The shipped `spec/files.json` still passes. ## Local review The pass raised one finding, declined on evidence. - `spec/validate.py`, the new tree-source comparison, `introduced`, design. The reviewer asked for one `paths_overlap` helper shared with the tree-target loop. Declined: no such helper exists to reuse, and the repeated expression is correct rather than a defect class. Moving the tree-target site is outside this change's claim, which is the `repoOwnedTrees` check only. Closes on promotion: #3528 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Copilot

🔵 Needs a closer look

The transient Copilot retry and review-gate changes depend on live GitHub Actions logs and GraphQL event ordering that mocked tests cannot fully validate, so real-run verification is needed before approval.

No open issues 🎉

1 resolved since last review

Reject repo-owned sources overlapping carried tree declarations

_{Review effort: Lite}

ptr727 · 2026-10-28T00:58:59Z

Answering the round's headline, "real-run verification is needed before approval", for the transient Copilot retry (#3542). I verified it against live runs from develop at e1c3a018, read-only.

Positive case, ptr727/HomeAutomation-Config#1430. Its refusal at 2026-10-26T19:22:25Z on 1e7291a3 is the failure Tell a Transient Copilot Fetch Error From the Quota Body Before Advising No Re-Request #3510 describes. Calling run_cause live with that commit and time finds failed Copilot run 40386072573 through the Actions REST API. It reads the job log and returns no error type, no rate-limit sentence, and the line HttpError: The request could not be processed because too many files changed. So transient_error gives that line, and the refusal reads as transient, not quota.
Fallback case, Promote the Env and Secret File Naming Rule to Main #3507. Its refusal at 2026-10-26T18:51:36Z has no failed Copilot run at or before it. The only Copilot run on that head succeeded at 18:21Z, and its log has no such line. run_cause returns None, and pr_review.py status 3507 prints RUN LOG READING: possible quota. The run log was unread or names neither a rate limit nor a transient fetch failure. That is the conservative reading the change keeps.
GraphQL reads. status ran cleanly against both live pull requests. The request-ordering state (owed, pending, sent, spent) is not exercised live, because neither refusal is still a pull request's newest review. The unit tests in tests/test_pr_review.py cover each state against constructed timelines.

Outcome: verified, no change needed.

ptr727 and others added 4 commits October 27, 2026 16:14

Copilot AI lite review requested due to automatic review settings October 28, 2026 00:16

Copilot started reviewing on behalf of ptr727 October 28, 2026 00:16

View session

Copilot AI reviewed Oct 28, 2026

View reviewed changes

Comment thread spec/validate.py

This was referenced Oct 28, 2026

Triage Financial-Modeling's tests/test_audit.py as a Path Collision, and Consider Keeping Hub tests/ Out of the Hub-Only Detector #3528

Closed

Refuse a Repo-Owned Tree Overlapping Carried Content #3545

Merged

ptr727 requested a lite review from Copilot October 28, 2026 00:49

Copilot started reviewing on behalf of ptr727 October 28, 2026 00:50

View session

Copilot AI reviewed Oct 28, 2026

View reviewed changes

ptr727 merged commit 65adbf1 into main Oct 28, 2026
14 checks passed

Navigation Menu

Repository navigation

Promote Develop: Resync-Filed Model Tiers, Handoff Lifecycle, Hub-Only Tests, and Copilot Transient Fixes - #3544

Promote Develop: Resync-Filed Model Tiers, Handoff Lifecycle, Hub-Only Tests, and Copilot Transient Fixes#3544
ptr727 merged 5 commits into
mainfrom
develop

ptr727 commented Oct 28, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Oct 28, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Oct 28, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

ptr727 commented Oct 28, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

ptr727 commented Oct 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

coderabbitai Bot commented Oct 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov Bot commented Oct 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

ptr727 commented Oct 28, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

ptr727 commented Oct 28, 2026 •

edited

Loading

coderabbitai Bot commented Oct 28, 2026 •

edited

Loading

codecov Bot commented Oct 28, 2026 •

edited

Loading