Pattern miner: Add threat-detection and job-discriminator tips from upstream mining

[github-actions]

Summary

Mined two small, evidence-backed refinements for existing curated archetypes. No new archetype was added this run — every recurring shape found that had two or more corroborating upstream examples was already represented in the library under a different tip, trigger, or safe-output recommendation.

Mining evidence

Sources mined:

• /tmp/gh-aw/data/upstream/githubnext-agentics/files/workflows/*.md (58 workflow files)
• /tmp/gh-aw/data/upstream/github-gh-aw/files/.github/aw/*.md (gh-aw v0.87+ docs)

1. threat-detection: false for code-pattern scanners (security-scanner)

• Upstream evidence: githubnext-agentics/files/workflows/malicious-code-scan.md (the direct upstream equivalent of the security-scanner archetype) and ai-moderator.md both explicitly set safe-outputs.threat-detection: false.
• Upstream docs corroborate this as a deliberate, recommended pattern rather than a one-off: .github/aw/workflow-patterns.md ("Release Automation Pattern") and .github/aw/release-workflow.md both instruct disabling threat-detection whenever generated output legitimately contains code snippets, which is exactly what a malicious/suspicious-code scanner's findings do (they must quote the suspicious code to be actionable).
• Gap: the library's security-scanner archetype tips covered scope, reporting target, noop, and dedup, but nothing about threat-detection false positives on the scanner's own quoted-code output.
• Change: appended one tip to patterns/archetypes/security-scanner.json.

2. concurrency.job-discriminator for schedule+workflow_dispatch archetypes (responsive-docs-tester)

• Upstream evidence: three workflows combining schedule with workflow_dispatch — multi-device-docs-tester.md (direct source for the responsive-docs-tester archetype), repo-assist.md, and vex-generator.md — all set concurrency.job-discriminator: ${{ github.event_name == 'schedule' && 'scheduled' || github.run_id }}.
• Upstream docs (.github/aw/syntax-core.md) document job-discriminator as an intentional feature specifically to prevent fan-out cancellations when "multiple workflow instances run concurrently with different inputs," listing the scheduled/dispatched-fallback expression as a named common usage.
• Gap: the library had no tip anywhere about job-discriminator, and responsive-docs-tester combines schedule with implicit manual re-runs for debugging a failing device/viewport, making it the best-matched existing archetype for this guidance.
• Change: appended one tip to patterns/archetypes/responsive-docs-tester.json.

What changed in patterns/

• patterns/archetypes/security-scanner.json: added one tips entry (no other field touched).
• patterns/archetypes/responsive-docs-tester.json: added one tips entry (no other field touched).
• patterns/manifest.json, patterns/workflow-generation.json: untouched (no new archetype).

Validation

• npm test: 329 tests passed across 19 test files (19/19 files green).
• npm run build: succeeded, dist/ produced cleanly including the two updated archetype JSON files.

Candidates considered and deliberately left out

• New archetype for wiki-to-code / wiki-driven workflows (agentic-wiki-coder.md, agentic-wiki-writer.md): only one upstream repo pair, no corroborating docs recommending gollum as a standalone archetype trigger. Below the two-workflow-or-docs bar.
• lock-for-agent, skip-roles, skip-bots, hide-comment, forks: moderation bundle (ai-moderator.md): all appear in a single upstream workflow; skip-roles/skip-bots are documented generically in syntax-core.md but not tied to a moderation-specific recommendation with a second corroborating example, so this stayed below the evidence bar for a refinement.
• group: true on create-issue for batched findings: three upstream workflows use it (discussion-task-miner.md, duplicate-code-detector.md, issue-arborist.md), but this exact tip ("use create-issue's group: true so they land as linked sub-issues") is already present verbatim in patterns/archetypes/issue-hierarchy-manager.json.
• Discussion-mining into issues (discussion-task-miner.md): a real recurring shape, but its core techniques (cache-memory history tracking, group: true, expires, bounded-window discussion scanning) are already spread across code-health-auditor, backlog-drip, and issue-hierarchy-manager tips; adding a new archetype would duplicate rather than add value within this run's "at most one archetype" budget.
• VEX/cost-tracker/sub-issue-closer style single-purpose workflows: each already has a matching curated archetype (vex-statement-generator, agent-cost-tracker, issue-hierarchy-manager).

Generated by Pattern Miner · copilot · auto · 395 AIC · ⌖ 10.3 AIC · ⊞ 8.3K · ◷