skill-validator

name	skill-validator
description	Validates skill profiles against the spec and explains how to fix failures before a PR is opened or merged.
owner	bcgov
tags	["meta","validation"]
metadata	{"internal":true}

Skill Validator

Use When

• The user wants to know whether a skill passes the spec.
• A PR check failed and the user needs to understand the error.
• The user is about to open a PR and wants a local pre-flight check.

Don't Use When

• The user wants to create a new skill from scratch → use skill-author.
• The failure is a Python/tooling error unrelated to a skill's contents.

Workflow

1. For one skill, run uv run python scripts/validate_skill.py skills/<name>/SKILL.md.
2. For everything, run uv run python scripts/validate_skill.py --all.
3. To mirror CI on a branch, run uv run python scripts/validate_skill.py --base origin/main.
4. Read each - <message> line under a failing file and map it to the offending section or field.
5. Edit the SKILL.md to fix it, then re-run until the file shows ✓.

Rules

• Always re-run the validator after each fix rather than batching guesses. (Why: one fix can reveal or mask another, and the validator is fast.)
• Never edit scripts/validate_skill.py to make a skill pass. (Why: the spec is the contract; loosening the validator weakens every skill in the catalogue.)

Examples

• "Does my skill pass?" → run the validator on its SKILL.md and report ✓ or the exact failures.
• "The PR check says 'missing required section Edge Cases'" → add a non-empty ## Edge Cases section and re-validate.
• "Validate everything before I push" → run --all and summarize the results.

Edge Cases

• If --base finds no changed skills → report that nothing changed; there is nothing to validate.
• If the error names a missing or empty section → add the section heading (verbatim) with at least one bullet of content and re-run.

References

See spec/SKILL_SPEC.md for what each check enforces and CONTRIBUTING.md for the PR-check summary.