docs(rules): a rejected flag reads as "no results"; CI waits race check registration

[laurigates]

Two silent-wrong-answer traps, both hit live this session while deduping an upstream report and merging a pair of PRs.

tool-use-patterns.md — a rejected flag looks exactly like "no results"

Any cmd … | jq/grep whose non-zero exit yields empty stdout masquerades as a legitimate empty result. The worst place for that is a dedup step: you conclude nobody reported the bug and file a duplicate.

Live instance: gh search issues --state all is invalid — that flag takes only {open|closed}; all belongs to gh issue list. It prints usage to stderr, so a --jq pipeline emits nothing. That produced six consecutive false "no duplicate" verdicts, caught by exactly one control run against a title I knew existed, which failed identically and exposed the bad flag.

The durable rule is the control-test discipline from never-fabricate-test-identifiers.md applied to search: control-test every negative that gates an action.

git-hazards.md #7 — waiting for CI races check registration

A loop on "zero pending checks" can exit immediately after a push or update-branch, because zero-pending is trivially true before the jobs are registered. Observed state=CLEAN on a single check while three CI jobs had not yet appeared — merging on that read merges untested code. Gate on both nothing-pending and a minimum expected check count.

Same root cause as the mergeability race directly above it in the file: an asynchronously-populated field read once, too early.

Budget bump — 110,000 → 114,000

The test file asks for a justification in the commit message, so: the always-loaded surface had reached 109,998 of 110,000 — two bytes free. At that point the budget had stopped being a cleanup trigger and become a hard stop on any new rule content.

Good-faith steps taken before bumping:

• both additions compressed ~50% from their first draft
• both host files brought back under PER_FILE_CAP_BYTES (9,917 and 9,706, from 10,835 and 10,147)

The cleanup pass is deferred, not skipped. The standing CONSOLIDATE candidate — multi-model-delegation.md (~6.6 kB), which largely duplicates the agent-patterns-plugin:multi-model-delegation skill — is recorded in the test file's comment so the budget can ratchet back toward 110,000 once it lands. It was being edited by a concurrent session at the time, so touching it here would have collided.

dot_zshrc.tmpl was modified in the working tree by other work and is deliberately not included.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PKbmsKBLRJMKNejkwujokZ

[github-actions]

⚠️ Broken Links Detected

Summary

Status	Count
🔍 Total	170
✅ Successful	160
⏳ Timeouts	0
🔀 Redirected	0
👻 Excluded	9
❓ Unknown	0
🚫 Errors	1

Errors per input

Errors in exact_dot_claude/docs/prds/daily-catchup.PRD.md

• [404] https://docs.anthropic.com/claude-code/commands | Failed: Network error: Not Found
  Full Github Actions output

---

Please fix the broken links before merging.