The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using Git to retrieve code diffs and commit logs for analysis.
The agent is directed to use git rev-parse to validate references, git diff to extract changes, and git log to examine commit history.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository's version history and external documentation, which could contain malicious instructions intended to influence the sub-agents' review behavior.
Ingestion points: Data enters the agent's context through git diff and git log output (containing author-controlled code and messages), as well as external specification files and issue tracker content fetched from docs/, specs/, or .scratch/.
Boundary markers: While the prompt for sub-agents includes formatting instructions (e.g., "Report... Under 400 words"), it lacks explicit delimiters or instructions to ignore embedded commands within the content being reviewed.
Capability inventory: The skill utilizes the agent's ability to execute shell commands (git) and read files within the repository.
Sanitization: There is no evidence of sanitization or filtering applied to the Git output or specification content before it is interpolated into the prompts for the parallel sub-agents.

code-review