Add evidence-based agent instructions audit skill 🤖🤖🤖

[fablerlabs]

Summary

• add agent-instructions-audit for evidence-based review of AGENTS.md, .github/copilot-instructions.md, and CLAUDE.md
• verify documented commands and paths against repository manifests, CI configuration, and the actual tree
• return a six-part score, worst-first fixes, and a verified-command table
• include the operating-principles field guide and its MIT license

Why this is distinct

This skill audits existing agent instructions against repository evidence. It does not generate another instructions file: its purpose is to catch stale commands, missing repository maps, unverifiable claims, conflicting instruction files, and absent verification loops, then provide concrete replacement text.

Affiliation

Submitted by Fabler Labs, the author and maintainer of the source skill. The source-repository metadata is included for provenance; the skill is self-contained and requires no external or paid service.

Validation

• bash eng/fix-line-endings.sh
• npm run skill:validate - all 372 skills valid
• Vally lint - 2/2 checks passed (spec-compliance, valid-refs)
• npm run build - generated the docs/README.skills.md install entry
• Copilot CLI 1.0.70 discovery smoke - copilot skill add and copilot skill list --json recognized the skill and bundled path
• git diff --check

[github-actions]

🟡 Contributor Reputation Check: MEDIUM risk

Check	Risk
Profile	MEDIUM
Credential audit	NONE

Maintainers: please review this contributor before merging.
See the workflow run for full details.
Automated check powered by AGT.