The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git diff to identify the scope of code changes for review. This is a standard development operation and is used appropriately for its stated purpose.
[DYNAMIC_EXECUTION]: The skill uses a template-based approach in references/reviewer-prompt.md to generate prompts for subagents. These subagents are restricted to read-only access (no writes, no git commands) during the review process.
[DATA_EXPOSURE]: The skill reads from ~/.config/pstack/pstack-models.md to determine which models to use for the review. This is a local configuration file for the pstack tool and does not involve sensitive credentials.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided code and diffs. It mitigates potential injection risks by using a structured review rubric and a 'Lead Judgment' phase where a primary agent filters and contextualizes findings, explicitly instructed to ignore noise and focus on correctness.

interrogate