| name | validate-agents-md |
| description | Validate AGENTS.md files for structural completeness, operational contract coverage, neutrality, and repository alignment. |
| version | 1.0.0 |
| allowed-tools | ["Read","Write"] |
| metadata | {"source_prompt":"prompts/skills/validate-agents-md.prompt.md","workflow_type":"validation"} |
Validate AGENTS.md
Purpose
Validate AGENTS.md files to ensure they are concise, operationally complete, and aligned with repository structure and governance.
When To Use
- Use this skill when adding or updating root or nested
AGENTS.md files.
- Use this skill before accepting AGENTS policy changes.
- Do not use this skill for user/project instruction validation.
How to Invoke
Load or attach this file's contents into your AI session to activate the workflow (paste, upload, or reference with #file:skills/validate-agents-md/SKILL.md in VS Code Copilot Chat). In Claude Projects, add it to project knowledge. See Invoking Skills in the README for full platform guidance.
Required Inputs
- Target
AGENTS.md path.
- Repository path context for link/path validation.
- Policy expectations for session-state and command namespace.
Workflow
- Run validation phases from
references/phase-checks.md.
- Check for
<!-- spec_version: ... --> comment: flag as WARNING if absent (pre-existing files may predate this requirement); flag as WARNING if present but older than 1.4.3.
- Generate report using
references/report-contract.md.
- Apply deterministic scoring from
references/scoring.md.
- Classify findings into critical/warning/enhancement buckets.
- Provide concrete fixes for operational contract gaps.
Output Expectations
- Structured markdown report with five validation phases.
- Compliance score and pass band.
- Explicit list of critical issues, warnings, and enhancements.
- Example fixes for common structural and policy failures.
- Report schema follows
references/report-contract.md.
- Scoring and grade bands follow
references/scoring.md.
Resources
- Phase checks:
references/phase-checks.md
- Report contract:
references/report-contract.md
- Scoring model:
references/scoring.md
- Example target:
../../AGENTS.md
Constraints And Safety
- Keep the scoring rubric deterministic; findings should be repository-specific and actionable.
- Avoid over-prescriptive style-only feedback.
- Preserve provider-neutral language.
- Do not duplicate full normative specs when suggesting fixes.