Pattern miner: Add pre-activation PR-cap tip to link-checker, linter-applier, dependency-monitor

[github-actions]

Upstream evidence

Mined githubnext-agentics and github-gh-aw (declared in data/import-sources.json) at /tmp/gh-aw/data/upstream/.

Recurring shape: a pre_activation gate that counts currently-open PRs (or issues) matching this workflow's title-prefix and skips the scheduled run once a cap is reached, to avoid spamming maintainers with too many open items from one automation.

11 upstream githubnext-agentics workflows demonstrate this shape with the identical MAX_OPEN_PRS check-step pattern:

• workflows/agentic-wiki-coder.md, workflows/ci-coach.md, workflows/doc-updater.md, workflows/efficiency-improver.md, workflows/glossary-maintainer.md, workflows/issue-monster.md, workflows/perf-improver.md, workflows/repo-assist.md, workflows/tech-content-editorial-board.md, workflows/test-improver.md, workflows/unbloat-docs.md

It is also documented explicitly in upstream docs as a first-class recommended pattern:

• github-gh-aw/files/.github/aw/workflow-patterns.md — "All You Can Eat Pattern" section: "Skipped runs are cheap. The check runs in the pre_activation job, so the agent never starts and no tokens are spent on a skipped tick." It also describes the skip-if-match cap shape (max: 1/max: N) for both issues and pull requests.

Why the existing library did not already cover this for these archetypes

The wizard's pattern library already carries this exact tip verbatim on code-improvement, documentation-updater, daily-test-improver, and performance-nut ("Add a pre-activation step that skips scheduled runs once too many open PRs share your title prefix, to avoid spamming maintainers"), confirming it is a known, valued tip. However it was missing from three other curated/measured archetypes that share the identical schedule → create-pull-request (or create-issue/create-pull-request) shape and are equally exposed to the same PR/issue pile-up risk:

• link-checker (schedule → pull-requests)
• linter-applier (schedule → pull-requests)
• dependency-monitor (schedule → issues, pull-requests)

What changed

Refinement only — appended the existing tip string to the tips array of:

• patterns/archetypes/link-checker.json
• patterns/archetypes/linter-applier.json
• patterns/archetypes/dependency-monitor.json

No changes to metadata, success_rate, count, top_repos, anti_patterns, trigger_combos, research_findings, degraded_workflows, recommended_triggers, recommended_safe_outputs, recommended_tools, or patterns/manifest.json / patterns/workflow-generation.json, per the change rules for a tips-only refinement. dependency-monitor has a measured success_rate/count, so only its tips array was touched, consistent with the rule to leave measured metadata untouched.

Validation

• npm test: 326/327 tests pass. The one failing test (test/copilot-instructions.test.js) asserts the committed .github/copilot-instructions.md "generated on" date matches manifest.metadata.generated_at; it fails identically on main before this change (verified with git stash), so it is pre-existing and unrelated to this diff.
• npm run build: succeeds, all three edited archetype JSON files build into dist/patterns/archetypes/.

Candidates deliberately left out

• New wiki-sync archetype (gollum-triggered wiki↔code sync, seen in agentic-wiki-coder.md/agentic-wiki-writer.md): only 2 upstream examples and no corroborating "recommended pattern" documentation beyond the general memory.md wiki-backend note — did not meet the two-source-plus-doc-or-two-source bar strongly enough to justify a new archetype in this run, and gollum trigger has no existing wizard trigger-rendering support to extend safely within scope.
• Discussion-based periodic report shape (weekly-issue-activity.md, repo-chronicle.md, weekly-research.md using create-discussion + close-older-discussions): already fully covered by the existing community-digest archetype and its tips.
• max: 1 safe-output volume caps generally (seen across many upstream workflows): already covered by existing "All You Can Eat" style tips on code-health-auditor, link-checker, and others (skip-if-match, expires, close-older variants); adding more would duplicate existing guidance rather than fill a gap.

Generated by Pattern Miner · copilot · auto · 131.7 AIC · ⌖ 10.7 AIC · ⊞ 8.3K · ◷

[github-actions]

Reviewer kind: auto → data-integrity review (JSON-only pattern-library change; no TS/UI surface touched by either persona).

Reviewed the diff: three tips array appends across patterns/archetypes/{dependency-monitor,link-checker,linter-applier}.json, using the exact tip string already present verbatim on code-improvement, daily-test-improver, documentation-updater, and performance-nut. Verified:

• All three JSON files remain valid and parse correctly.
• No duplicate tips introduced within any file.
• No changes to success_rate, count, metadata, anti_patterns, trigger_combos, or patterns/manifest.json / workflow-generation.json, consistent with a tips-only refinement for archetypes with measured data (dependency-monitor).
• PR description's claim that 4 other archetypes already carry this tip checks out.

No blocking issues found. Small, well-scoped, low-risk data refinement.

Generated by Specialist PR Review for #292 · copilot · auto · 17.2 AIC · ⌖ 4.85 AIC · ⊞ 6.8K