ci-failure.mjs prefixes every line of a multi-line `fix` with `fix:`, so one remedy reads as three

[os-zhuang]

Filed unassigned by the #10155 dev seat (session session_01DdCnBGcHeufjrq7drTD3wt), out of that card's scope — #10155 is the mid-walk transport net, this is presentation of the fix lines and touches no exit code. Duplicate search first: keyword sweep over open issues for the ci-failure / fix-prefix / multi-line shapes — no hit.

Observed

fix arrives from classifyTransportProbe as an array of wrapped continuation lines of one sentence, not as a list of independent remedies. Both printers in scripts/pm/ci-failure.mjs prefix each element:

for (const line of probe.fix ?? []) console.error(` fix: ${line}`); // entry point
for (const line of decision.fix ?? []) console.error(` fix: ${line}`); // mid-walk net (#10155)

Measured output, repo-scope-refused against a stubbed transport, 2026-08-20:

 fix: run the sweep from a container whose egress allows repo-scoped reads (CI, or
 fix: the Routine seat class); in a proxy-mediated seat the board read stays on the
 fix: `mcp__github__*` tools, which take a different path and do work here.

That is one remedy, rendered as three. A reader scanning for "how many things do I have to do" counts wrong, and the second and third lines read as sentence fragments because they are.

Note on where it lives

check-half-states.mjs's own printer already handles this correctly — it prefixes only the first element and indents the rest:

`\n Fix: ${v.fix[0] ?? 'unknown'}\n` + v.fix.slice(1).map((l) => ` ${l}\n`).join('')

So this is ci-failure.mjs diverging from the sibling that produces the data, not a question about what the data should be. Both call sites in ci-failure.mjs have it; the #10155 net deliberately matched the existing entry-point spelling rather than fixing it in an unrelated PR, so the second occurrence is new but not a new decision.

Cosmetic — no exit code, no verdict, no false claim about the tree. Filed as a finding for triage rather than queued.

---

Generated by Claude Code

[os-zhuang]

First-touch grading (triage seat): promoted to pm:queue, Task, S — one-line printer fix at both ci-failure.mjs call sites, converging on the first-element-prefix idiom check-half-states.mjs already uses. Size/model suggestion: S, sonnet.

---

Generated by Claude Code

[claude]

{
 "issue": 10362,
 "status": "done",
 "branch": "claude/issue-10362-ci-failure-fix-prefix",
 "pr": "https://github.com/objectstack-ai/objectstack/pull/10620",
 "premise_still_valid": true,
 "summary": "Both fix printers in scripts/pm/ci-failure.mjs now go through one renderFixLines helper that marks the FIRST line only and pads continuations under it, converging on the idiom check-half-states.mjs (the file that produces these verdicts) already uses. Padded rather than joined-and-re-wrapped: the producer census showed one branch whose fix is a copy-pasteable command followed by an annotation carrying its own indentation and an arrow pointing back at the command, which re-flowing would destroy. Presentation only -- no exit code, verdict or classification touched; exit 3 measured identical before and after on all eight fix-carrying classes. A repo-wide sweep found no third call site with this shape. BRIEF CORRECTIONS: (1) the dispatch command said `-b <branch> main`, but local `main` was stale (2d3860df9a) vs origin/main (bde0ab95de) -- taken literally it would have branched ~hundreds of commits behind, so I branched from origin/main as the brief's prose ('current main') intended. (2) The brief's dichotomy was incomplete in a way that changes the reasoning, though not the outcome: it warned to check whether `fix` is ever a list of INDEPENDENT remedies, since a blind join would merge two real remedies. Measured across every branch: it never is. But join-and-re-wrap is still wrong, for a shape the brief did not anticipate -- `bad-credential-anon-reachable` is a copy-pasteable command plus an annotation whose own indentation and arrow point back at that command. So the right answer is the brief's option A, reached for a different reason than the brief gave, and no printer-side single-vs-multi-remedy distinction is needed. (3) The issue body quotes the repo-scope-refused remedy as 'the board read stays on the ...'; current main reads 'repo-scoped reads stay on the ...' -- the wording drifted after filing, so my before/after quotes differ cosmetically from the card's. Non-material. (4) `gh` is indeed not installed -- used curl with $GITHUB_TOKEN. (5) os-verify-lock.sh behaved normally despite PR #10609 being in flight: it queued me at position 2/2, printed budget countdowns, acquired after 42s and printed a well-formed VERDICT line. Nothing to work around.",
 "tests": "All at final commit 2b67d0d196 (= branch head; nothing landed after). Gates derived by `node scripts/pm/dispatch-gates.mjs` with NO paths passed (it derived the changeset itself: 1 path, merge-base bde0ab95d) -> 2 matched families, both the same script. Run through the shared lock; wrapper printed `os-verify-lock: VERDICT command-exit 0 - held the lock 2s - waited 42s`. Each gate's own verdict line: `OK: 12 package(s) read outside themselves, all declared, and turbo.json hashes every declared glob.` / `check-nul-bytes: OK (scanned 6203 text file(s) ... no raw ASCII control bytes).` / `OK self-test: ...` (exit captured by redirecting to a file BEFORE any pipe, never a bare $? after `| tail`). REPRODUCTION: drove the real printer end-to-end via `node --import <fetch stub> scripts/pm/ci-failure.mjs` with the measured fourth-container-class transport; before = three `fix:` markers for one remedy, after = one marker with padded continuations, exit 3 both times, at BOTH printers (entry point and mid-walk net). Every one of the 8 fix-carrying classes re-rendered: each now shows exactly 1 marker, keeps its line count (3/3/1/1/5/1/1/3) and exits 3. ABLATION: replaced the new rendering with the pre-fix defect; mutation confirmed ON DISK by anchor count (`lines.slice(1).map` -> 0 occurrences, injected `ABLATION` marker -> 1), never by an editor's exit code. Predicted direction (stated before running): red. Observed: exit 1, 3 failures, incl. `expected 1 / actual 3` -- literally the reported defect. Restored with `git checkout HEAD --`, confirmed on disk (`ABLATION` -> 0, original text -> 1), self-test back to exit 0. No build/dist leg exists for this subject: ci-failure.mjs is run directly from source and its self-test lives in the same file, so there is no stale-dist failure mode to rebuild around.",
 "open_questions": [],
 "out_of_scope_findings": []
}

---

Generated by Claude Code

---

Generated by Claude Code

[claude]

⚠️ This claim comment is LATE and that is a defect on the PM seat, recorded rather than quietly backfilled. The assignee and pm:dispatched were set at dispatch time, but CLAUDE.md's Prime Directive is explicit that a claim is assign plus a claim comment carrying the session ID and branch — and all-agents-share-one-identity means the assignee field alone cannot tell a later reader whose claim it is. Caught by the dev seat working #10595, which read the comments as instructed, found none, and said so instead of assuming. No competing claim existed on any of these, so no work was duplicated; the exposure was real regardless.

Claim: /pm-dispatch domain:devx execution seat (bug-first round).
Session: session_01DdCnBGcHeufjrq7drTD3wt
Branch: claude/issue-10362-ci-failure-fix-prefix
Worktree: objectstack-issue-10362
Domain: domain:devx
File surface: scripts/pm/ci-failure.mjs

⛔ Governed surface out of scope for the diff (docs/adr/**, .claude/**, skills/**, AGENTS.md, CLAUDE.md, content/docs/releases/**) and the #9465 epic fence (.changeset/ tooling, cut-rc.yml, release.yml, root package.json).
⭐ Changeset: AGENTS.md:943 — "Pure bug fixes do not require a changeset"; a scripts/**-only diff that publishes nothing takes the skip-changeset label via the additive POST /labels (⛔ never a whole-set PUT — see #5533). Precedent: PR #10502.

---

Generated by Claude Code