The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard git commands (git rev-parse, git log, git diff) to determine commit ranges and extract code changes. These commands are executed locally within the repository and are necessary for the skill's intended purpose of code review.
[PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it processes untrusted data (the code being reviewed) through git diff output.
Ingestion points: The {BASE_SHA}..{HEAD_SHA} diff output is ingested into the code-reviewer subagent context via the code-reviewer.md template.
Boundary markers: No explicit delimiters or instructions to ignore embedded prompts in the code are present in the template.
Capability inventory: The skill uses git commands and dispatches a subagent; it does not perform network operations or unsafe file writes.
Sanitization: No sanitization of the diff output is performed.
Risk Assessment: The risk is assessed as LOW because the reviewer agent is provided with a highly structured checklist and specific output format, which naturally constrains its behavior despite the ingestion of untrusted code comments.
[DATA_EXFILTRATION]: No network operations or access to sensitive credentials (SSH keys, AWS tokens) were detected. The SHAs used are public identifiers within the git environment.
[EXTERNAL_DOWNLOADS]: The skill does not download external scripts or packages from remote sources.

requesting-code-review