The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides detailed instructions for using the GitHub CLI (gh) to trigger workflows, monitor runs, and debug failures. It also suggests running local build tools such as pnpm and cargo to address issues found in the CI process.
[EXTERNAL_DOWNLOADS]: The agent is instructed to fetch workflow logs and build artifacts from the vendor's repository (codervisor/lean-spec) on GitHub using gh run view --log and gh run download. These downloads originate from a well-known and trusted service.
[PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection (Category 8) by processing untrusted data from the GitHub Actions environment.
Ingestion points: External data is ingested through workflow logs and build artifacts as described in SKILL.md and references/COMMANDS.md.
Boundary markers: The instructions lack delimiters or explicit guidance to treat retrieved content strictly as data, which may allow instructions embedded in logs or artifacts to influence the agent.
Capability inventory: The agent possesses capabilities to trigger workflows, download files, and execute local development scripts (pnpm, cargo).
Sanitization: There is no evidence of sanitization or content validation for the data retrieved from the GitHub Actions environment before it is analyzed by the agent.

github-actions