The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git commands (e.g., git diff, git rev-parse) to determine the scope of code changes for review.
[COMMAND_EXECUTION]: Executes external AI CLI tools, including Anthropic's claude, OpenAI's codex, and cursor-agent, via the scripts/run_review_agent.sh wrapper.
[COMMAND_EXECUTION]: Performs environmental probes to check for tool availability and authentication status using commands like command -v and vendor-specific status checks (e.g., claude auth status).
[PROMPT_INJECTION]: The skill identifies as a surface for indirect prompt injection because it processes untrusted data (code diffs and user-provided standards) and interpolates them into sub-agent prompts.
Ingestion points: The skill reads repository diffs into tmp/review/<feature>/changes.diff and ingests user-provided review standard documents.
Boundary markers: It uses explicit headers and section markers (e.g., --- REVIEWER IDENTITY ---) to separate instructions from untrusted data in the generated prompts.
Capability inventory: The skill can write files to the repository, execute shell commands, and invoke external AI sub-agents.
Sanitization: Inputs for file paths and reviewer IDs are validated against a restricted character set regex (^[A-Za-z0-9._-]+$).

code-review