[turtlemonvh]

Decision needed: none. Once this merges, close #121 (the retired summary issue). Proposed: close it as "superseded by the local report".

What changed

• .claude/night-crew.json: summary_issue: 121 replaced with report_path: "STATUS_REPORT.html". .gitignore now ignores /STATUS_REPORT.html.
• CONTRIBUTORS.md and CLAUDE.md: the passages about the morning-summary comment on night-crew: morning summary log #121 now describe the local report file (gitignored, and shared as an Artifact only when asked). CLAUDE.md still says the core plugin injects these rules.
• blanket-ready-batch: removed "Merge PRs" from the "does not do" list and "never auto-merge" from the description, because step 6 already merges ship-to-merge + risk:low PRs. The subagent brief now gives a PR-body template in night-crew's "PR shape and size" order, plus the ~400-line budget and the 🤖 Claude: comment rule.
• Comment marker: CONTRIBUTORS.md "Visible attribution" and the blanket-issue-triage templates now start each comment with a 🤖 Claude: line instead of **🤖 Claude** placed after the hidden label-note.

Judgment calls

• I widened the marker change beyond ready-batch. night-crew treats any comment that doesn't start with 🤖 Claude: as the reviewer speaking. The old triage notes started with <!-- blanket-label-note -->, so night-crew could misread them. ready-batch posts its comments through triage, so changing ready-batch alone would not have fixed this. The audit skill matches the hidden marker anywhere in a comment, so the new line order doesn't affect it. Old comments are not backfilled.
• I gitignored the report in .gitignore. night-crew's own fallback is .git/info/exclude, and its loop never edits .gitignore. A committed rule covers every checkout without that step. The .claude/* block already sets the precedent for committed tooling ignores.

Verification

• No Go, script or CI files changed, so I didn't run the docker test suites or make check-fmt.
• python3 -m json.tool .claude/night-crew.json succeeds.
• git check-ignore -v STATUS_REPORT.html matches the new .gitignore:32 rule.
• No summary_issue or #121 references are left outside the one historical note in CONTRIBUTORS.md.

Closes #222

🤖 Generated with Claude Code

[turtlemonvh]

I gitignored the report in .gitignore. night-crew's own fallback is .git/info/exclude, and its loop never edits .gitignore. A committed rule covers every checkout without that step. The .claude/* block already sets the precedent for committed tooling ignores.

I don't understand this.
What does "night-crew's own fallback" mean? What is it a "fallback" from?
What told you "its loop never edits .gitignore"?
When you say "A committed rule covers every checkout without that step." - I understand you mean a .gitignore rule, but what is "that step"?
What is "The .claude/* block"?

Aside from that, the rest LGTM

[turtlemonvh]

🤖 Claude: That paragraph explained it badly. Plain answers:

• "night-crew's own fallback": night-crew's setup step (core 0.12.0, skills/night-crew/SKILL.md, step 3) runs git check-ignore -q STATUS_REPORT.html. If the report isn't already ignored, it appends the path to .git/info/exclude. That file is local to one checkout and never committed. "Fallback" meant "what it does when the path isn't ignored yet."
• "its loop never edits .gitignore": the same step says, word for word, "Never git add the report and never edit the repo's .gitignore for it."
• "that step": the append to .git/info/exclude, which every fresh checkout has to repeat for itself. A committed .gitignore rule makes it unnecessary.
• "the .claude/* block": lines 25–30 of .gitignore. They ignore .claude/* except the committed skills and night-crew.json, which is the existing example of a tooling ignore living in .gitignore.

Where the line came from: my instructions to the subagent asked it to "make sure STATUS_REPORT.html is gitignored". night-crew didn't ask for it, and it doesn't need it, since night-crew already keeps the file untracked on its own.

Recommendation: drop the /STATUS_REPORT.html line from .gitignore and leave ignoring to night-crew. The cost is that someone opening the repo without night-crew could see an untracked STATUS_REPORT.html after a run on their machine. The alternative is to keep the line, which is harmless but goes against night-crew's own rule. I'll push the removal unless you'd rather keep it.

[turtlemonvh]

OK it's a fair point, but yes, let's remove the .gitignore rule