The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository being reviewed, such as code files and git history, to identify and fix issues. This creates a surface where malicious instructions embedded in the repository code could influence the agent's actions.
Ingestion points: Target repository files, git diff, git show, and git log output (SKILL.md, Steps 1, 2, and 3).
Boundary markers: The skill does not define explicit delimiters or instructions to treat the analyzed code as data rather than instructions.
Capability inventory: The agent can modify files (Fix step) and execute arbitrary shell commands via the repository's test suite (Verify step).
Sanitization: There is no evidence of sanitization or filtering applied to the ingested code.
[COMMAND_EXECUTION]: The skill executes shell commands to interact with the version control system and project environment.
Evidence: Uses git diff, git show, and git log to analyze the code state (SKILL.md, Step 1 & 2). It also executes repository-specific test runners like npm test (references/report.md) and other checks (SKILL.md, Step 5.2).
[DYNAMIC_EXECUTION]: The skill performs dynamic execution of repository code during the verification process.
Evidence: The Verify step (SKILL.md, Step 5.2) requires the agent to run the repository's tests to confirm that fixes work as intended.

zen-review