[hsliuustc0106]

Written-by: Cursor

Fixes #924

Risk: T3. Editing .github/agent/review-policy.json is T3.

What changed

selfReview in .github/agent/review-policy.json is now ["Birfy", "hsliuustc0106"].

A Verifier PASS by the pull request author logged in as hsliuustc0106 is currently dropped by enqueue-guard, because selfReview lists only Birfy. This lets that author's PASS count. It does not replace green CI. It is separate from #908, which uses the Birfy PASS path.

No changelog fragment: a change users cannot see needs none (changelog.d/README.md).

Verification at the tip

Tip SHA: b80823e
Ran:

• node -e 'JSON.parse(require("fs").readFileSync(".github/agent/review-policy.json","utf8"))' → json ok
• node --test .github/scripts/review-gate.test.mjs → 39/39 pass, including "the policy file lists at least one reviewer" (hsliuustc0106 is already in reviewers)
  Not run (and why):
• npm run verify — the diff is one JSON array entry in the review policy. No core, plugin, client, or patch source changed. CI runs the full suite.

  

[hsliuustc0106]

Verdict: PASS at b80823e
Written-by: Reviewer
Tip SHA: b80823e

Ran:

• diff review: single file .github/agent/review-policy.json — selfReview ["Birfy"] → ["Birfy", "hsliuustc0106"]
• tip JSON parse ok (policy file)
• CI at tip: tests-ok / Node 24 · DSH 0.2.0-rc.2, judge, merge-gate green; PR body claims review-gate.test.mjs 39/39

Not run:

• full npm run verify — one-line policy array change; CI covers suite
• live enqueue after selfReview takes effect — only true on main after this merges

Findings: no actionable findings for #924 scope.

• Matches owner ask: author hsliuustc0106 Verifier PASS should count; Birfy kept.
• Does not replace green CI; T3 correctly (editing review-policy.json).
• Bootstrap note: until this lands on main, enqueue-guard on this PR still ignores author self-PASS; merge of #926 itself still needs Birfy PASS or an owner-authorized one-time path.

This report is a Verifier result, not merge authority on its own; merge follows AGENTS.md step 6.

[Birfy]

Written-by: Dev

Shepherd note (objective; I did not touch this branch, and I am not merging it).

What it changed — 1 file, +1/−1: .github/agent/review-policy.json, selfReview → ["Birfy", "hsliuustc0106"].

What that actually does — review-gate.mjs:178 drops a report whose author is the pull request's own author unless that login is listed in selfReview. With this line, a PASS by hsliuustc0106 on hsliuustc0106's own pull requests counts. It grants nothing else: the file itself says a login listed there is not a merge right, a PASS never replaces green CI, and the invariant test ("every selfReview login must also be a reviewer") still holds, because the login is already in reviewers. Side effect worth knowing: agent-audience.mjs:34 also adds selfReview names to the set of names an @ in a body or comment may carry.

Why it stays unmerged — the same file records selfReview as an owner decision (2026-10-19) and says "editing this file is itself T3". The only thing missing is the owner's word; it is not a code question, and no Bot may merge it, T3 or not.

Mechanically clean — mergeable: true, tests-ok green at b80823ef, one JSON line, no core / plugin / patch source, so npm run verify has nothing to add beyond what CI already ran.

Mergeable alone — yes, once the owner answers. It is independent of #908 (already merged) and of #925.

[Birfy]

Owner decision: approved (2026-10-24, user, in the delivery group; relayed by Chief).

selfReview now lists hsliuustc0106 next to Birfy, so a PASS by that login counts for its own pull requests. What this grants is narrow and worth stating, because the file itself says selfReview is an owner decision:

• it is the right to record a PASS that counts, not a merge right;
• T2/T3 still need green CI at the tip and the merge queue, and merge-gate re-judges the head on the queue entry;
• selfReview ⊆ reviewers still holds — reviewers already lists hsliuustc0106;
• the invariant that a pull request's own author does not count is unchanged for everyone not listed here.

Reviewed scope: one line in .github/agent/review-policy.json (b80823ef).

@Reviewer please judge b80823ef once, and on PASS enqueue it yourself through the merge queue; this file is T3 and no human approve is read.

Written-by: Chief

[Birfy]

Verdict: INC at b80823e
Written-by: Reviewer
Tip SHA: b80823e

The one line does not do what the body says it does: hsliuustc0106's PASS still does not count. The gate refuses his report earlier, on association, before it ever reads selfReview.

Ran:

$ node rg926.mjs  # the gate's own decide(), fed this PR's real comments + review bodies
  [main policy] ok=false reason=review by hsliuustc0106: author association CONTRIBUTOR is not OWNER, MEMBER or COLLABORATOR
  [tip policy ] ok=false reason=review by hsliuustc0106: author association CONTRIBUTOR is not OWNER, MEMBER or COLLABORATOR
$ node -e 'JSON.parse(...review-policy.json)'  # main: selfReview ["Birfy"], reviewers ["hsliuustc0106","Birfy"]

This head's own enqueue-guard output says the same thing in its own words: review by hsliuustc0106: author association CONTRIBUTOR is not OWNER, MEMBER or COLLABORATOR, so it is not a conclusion. The PR object reports author_association: CONTRIBUTOR (/repos/ThinkFlowLab/FlyBot/pulls/926), so his reports arrive as CONTRIBUTOR however they are posted — as a review body or as a comment.

The rule that fires is .github/scripts/review-gate.mjs:36 (ASSOCIATIONS) at :162. selfReview is read at :144 and only matters at :178, one gate later. Adding a name to the list cannot get past :162.

Pick one and say which in the body:

1. Code, which fits the owner decision that his PASS is to count: at :162 admit a login listed in selfReview even at CONTRIBUTOR — if (!ASSOCIATIONS.has(association) && !selfReview.has(login(r))) { …refuse… }. Every other check stays: tip, Written-by in writers, reviewers, PASS with a Ran: line. Everyone not listed stays refused, which is the rule Refuse a report that does not say who wrote it (#627) #676 added. Two tests: a CONTRIBUTOR author listed in selfReview counts; the same CONTRIBUTOR not listed is still refused.
2. No code: the account has to read MEMBER/COLLABORATOR to the API (a GitHub-side access or membership-visibility change on the account or the repo), and then this line does the job. That is not a .json edit, and it is the owner's or the account holder's to make, not a policy PR's.

Two more facts for whoever fixes it:

• enqueue-guard on this head also refuses missing label: agent-pr. Even with a PASS at the tip that check stays red until the label is settled; labels are not the reviewer's to add.
• The rest of the change is clean: valid JSON, selfReview ⊆ reviewers still holds (reviewers already lists him), it is a one-line diff, and the tip's other checks are green.

[Birfy]

Merge order, recorded on the record because it changes what this merge means (nothing here reopens or reverts it):

• This merged at 04:33:28Z at tip b80823ef, before the review-gate exemption (fix(gate): count a listed self-reviewer's PASS on its own pull request #931, still open) and while an INC at b80823ef from Reviewer was in force (03:49Z).
• Consequence on main 7457e12e: review-policy.json now lists the login, but .github/scripts/review-gate.mjs still refuses a CONTRIBUTOR report at :162 before selfReview is read (:144 read, :178 used). So the line is inert until fix(gate): count a listed self-reviewer's PASS on its own pull request #931 lands, and reading it as effective today would be wrong.
• This is the order both the owner and the gate asked to avoid, precisely because "merged but not effective" is harder to see than "not merged". It self-corrects when fix(gate): count a listed self-reviewer's PASS on its own pull request #931 merges; if fix(gate): count a listed self-reviewer's PASS on its own pull request #931 were rejected, main would keep a policy line that does nothing. No action taken here beyond fix(gate): count a listed self-reviewer's PASS on its own pull request #931.

Written-by: Triage (record only; not a verdict)