Pattern miner: Add mentions/backlinks suppression tip to status-report and community-digest

[github-actions]

Mining summary

Mined upstream sources under data/import-sources.json: githubnext-agentics (workflow files) and github-gh-aw (.github/aw/*.md reusable guidance).

Recurring pattern found

Six upstream githubnext-agentics workflows that periodically read broad repository activity and publish it as a new issue or discussion all explicitly set mentions: false and allowed-github-references: [] in safe-outputs:

• workflows/repo-status.md
• workflows/team-status.md
• workflows/plan.md
• workflows/tech-content-editorial-board.md
• workflows/accessibility-review.md
• workflows/adhoc-qa.md

This is corroborated by upstream documentation: github-gh-aw/files/.github/aw/report.md, section "Avoiding Mentions and Backlinks" (lines ~140-152), which explicitly recommends this combination — plus max-bot-mentions: 0 — for exactly this shape of workflow: "Without filtering, @username notifies users and #123 creates backlinks every run."

Both conditions for a candidate pattern are met: at least two (here, six) upstream workflows demonstrate the shape, and upstream documentation independently describes it as recommended practice.

Why the existing library didn't already cover it

I checked /tmp/gh-aw/data/current-library.json, patterns/manifest.json, and every existing patterns/archetypes/*.json tip for mentions, allowed-github-references, backlink, and notification — none appear anywhere in the current pattern library. The status-report and community-digest archetypes (the two library entries matching this "periodic broad-activity report" shape) had no guidance on suppressing unwanted @mention notifications or issue/PR backlinks from generated reports.

What changed

Added one tip each (refinement only, no new archetype, no other fields touched):

• patterns/archetypes/status-report.json: added tip "Set mentions: false, allowed-github-references: [], and max-bot-mentions: 0 on safe-outputs so periodic reports don't notify users or create backlinks/bot-trigger side effects every run"
• patterns/archetypes/community-digest.json: added the equivalent tip for digests

No other fields (metadata, anti_patterns, trigger_combos, research_findings, degraded_workflows, success_rate, count, top_repos) were modified. patterns/manifest.json and patterns/workflow-generation.json were left untouched since this is a tip-only refinement to already-existing curated archetype entries, and the new tip already surfaces into generated prompts via the existing actionableWorkflowTips() pass-through in src/js/workflow.js (it does not match any of that function's exclusion patterns).

Validation

• npm test: 326/327 passing. The 1 failing test (test/copilot-instructions.test.js) fails identically on main before this change (confirmed via git stash) — it compares a hardcoded date string in the custom-instructions system prompt against patterns/manifest.json's metadata.generated_at, which is a pre-existing drift unrelated to this change.
• npm run build: succeeds, all pattern JSON files (including the two modified) build correctly into dist/.

Candidates deliberately left out

• skip-roles / skip-bots / skip-author-associations (seen in ai-moderator.md, several pre_activation gating steps): recurring, but already effectively covered by the existing trigger-combo library (schedule+skip-bots, pull_request+skip-bots are tracked combos) and the pre-activation PR-cap tips already present on several archetypes.
• hide-comment safe output (seen once in ai-moderator.md, documented in safe-outputs-automation.md/designer-mappings.md): only one upstream workflow demonstrates it; doesn't meet the "at least two workflows" bar despite documentation support, and content-moderation already recommends add-labels+add-comment without hiding.
• upload-asset + Python chart generation (seen in weekly-issue-activity.md, repo-chronicle.md, and documented across charts.md/charts-trending.md/asciicharts.md): a real recurring pattern, but it's a broader cross-cutting capability (chart generation for any report-style archetype) rather than a single archetype-scoped tip, and would need its own extras entry plus generation-logic wiring in src/js/workflow.js to render correctly — too large a change to bundle safely into this run. Left for a future dedicated mining pass.
• repo-memory on gollum (wiki edit) triggers (agentic-wiki-coder.md, agentic-wiki-writer.md): interesting but gollum isn't a documented supported trigger anywhere in the mined .github/aw/*.md guidance files, so I could not corroborate it against v0.87+ documentation as required.
• New "Discussion Task Miner" / "Sub-Issue Closer" archetype candidates: each is only demonstrated by a single upstream workflow (discussion-task-miner.md, sub-issue-closer.md respectively) with no second corroborating workflow or explicit "recommended pattern" documentation, so neither meets the two-workflow-or-documented-recommendation bar for a new archetype.

Generated by Pattern Miner · copilot · auto · 166.7 AIC · ⌖ 12.5 AIC · ⊞ 8.3K · ◷