The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for the agent to ingest and analyze GitHub Actions logs and workflow run artifacts, which are untrusted external inputs.\n
Ingestion points: gh aw logs and gh aw audit commands used in SKILL.md to fetch run data.\n
Boundary markers: The skill explicitly mandates a "read-only security-review agent per revision" and an "annotated overview of changed authored .md sources" to explain security impacts.\n
Capability inventory: The agent is guided to use gh aw compile, gh aw run, and gh run download to interact with the repository and workflows.\n
Sanitization: The instructions require the agent to apply "untrusted-evidence rules" and "credential triage" protocols to minimize the risk of executing or trusting malicious content within logs.\n- [COMMAND_EXECUTION]: The skill utilizes the gh (GitHub CLI) tool and the gh-aw extension for workflow management.\n
Evidence: Examples include gh aw logs, gh aw audit, and gh aw compile, along with standard utilities like jq and cat for log processing. These are standard developer tools and their use is consistent with the skill's purpose.

debugging-workflows