The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git diff to identify changes within the repository. This is a standard and expected operation for a tool focused on code review.
[SAFE]: Subagents are configured as readonly: true, which effectively sandboxes their execution and prevents them from modifying the local environment or repository during the review process.
[PROMPT_INJECTION]: The skill processes untrusted code content through the {DIFF_OR_FILES} variable in references/reviewer-prompt.md. This creates a surface for indirect prompt injection, but the risk is managed by the skill's multi-model consensus and lead-reviewer synthesis logic. Ingestion points: {DIFF_OR_FILES} in the prompt template. Boundary markers: Markdown headers are used. Capability inventory: Read-only repository access and subagent creation. Sanitization: None.

interrogate