The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands including git diff to extract code changes and gh issue comment to post review results to GitHub. These commands are standard for developer-oriented workflows.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and reviews untrusted data (source code changes), which creates a potential surface for indirect prompt injection.\n
Ingestion points: External code is ingested through git diff commands described in SKILL.md.\n
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the content being reviewed.\n
Capability inventory: The skill has network access through the GitHub CLI and can execute local subagents using codex-subagent.\n
Sanitization: There is no evidence of filtering or sanitizing code comments or strings within the diff before analysis.

comprehensive-review