The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses local shell commands and the GitHub CLI (gh) to automate git operations and PR management.
Evidence: Found throughout SKILL.md and the supporting rules files, using commands like git diff, git push, gh pr create, and gh pr checks --watch to manipulate the repository state.
Context: These operations are consistent with the skill's primary purpose as a PR automation utility.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its ingestion of untrusted external data during its triage and fix workflows.
Ingestion points: The skill reads GitHub Actions logs via gh run view --log-failed (SKILL.md Step 8) and processes Claude review comments via the implement-suggestion skill (rules/review-mode.md).
Capability inventory: The agent has significant capabilities, including modifying the local filesystem, pushing code to remote branches, and triggering CI reruns.
Boundary markers: While the skill instructs subagents to produce structured reports, it lacks explicit boundary markers or instructions to ignore potential directives embedded within the logs or comments being processed.
Sanitization: No explicit sanitization or filtering of external content is documented before it is passed to subagents for analysis or implementation.

create-pr