The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the GitHub CLI (gh) to interact with repository state, pull requests, and CI workflows.
Evidence includes the use of gh pr view, gh pr checks, gh run view, gh run cancel, and gh run rerun to manage and diagnose CI jobs.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources such as CI logs, PR metadata, and external check links.
Ingestion points: Untrusted data enters the agent context via gh run view --log-failed, gh pr checks, and content extracted from external check links (SKILL.md, steps 3-6).
Boundary markers: The SKILL.md contains an explicit 'External Side Effects' warning that PR metadata, commit messages, and logs must be treated as untrusted text and that embedded instructions should never be followed.
Capability inventory: The skill has the capability to modify local source code, commit changes, push to remote repositories, and cancel/rerun CI workflows.
Sanitization: The skill relies on natural language instructions to the agent to treat external data as diagnostic evidence only, rather than programmatic sanitization.
[DYNAMIC_EXECUTION]: The skill instructs the agent to manually load and follow instructions from other skill definitions found on the local filesystem.
Evidence: The 'Fix Ci procedure' section in SKILL.md directs the agent to 'Run another skill by reading its sibling ..//SKILL.md and following it'.

fix-ci