The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing local validation and packaging scripts within a CI/CD environment. This is the primary intended function for automating skill delivery.
Evidence: SKILL.md and references/skill_validation_and_packaging_ci.md include commands such as python3 ~/.claude/skills/skill-creator/scripts/quick_validate.py and python3 ~/.claude/skills/skill-creator/scripts/package_skill.py to gate and bundle skills.
[EXTERNAL_DOWNLOADS]: The skill references the installation of common, well-known utility packages and the use of standard GitHub Actions for pipeline automation.
Evidence: references/github_actions_and_build.md suggests using actions/checkout@v4 and actions/setup-python@v5, as well as running pip install pyyaml to support the validator script.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a validator and packager for other skills, creating an attack surface where malicious content in a target repository could attempt to influence the agent's behavior during the CI process.
Ingestion points: The skill reads SKILL.md frontmatter, reference files, and directory structures from external skill repositories as described in references/skill_validation_and_packaging_ci.md.
Boundary markers: The skill explicitly incorporates quick_validate.py to enforce structural integrity and YAML validity, which acts as a technical boundary.
Capability inventory: The pipeline has the capability to execute shell scripts, compile artifacts (.skill bundles), and upload assets to GitHub Releases or internal registries.
Sanitization: The skill advocates for human review (CODEOWNERS), YAML schema validation, and description length constraints to mitigate the risk of malicious metadata.

cicd-engineer