The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code diffs and file contents which could contain malicious instructions designed to influence or override the instructions of the LLM reviewers.
Ingestion points: Code diffs (obtained via git diff), file contents, commit messages, and PR descriptions are gathered in Step 1 of SKILL.md and passed to subagents.
Boundary markers: The prompt template in references/reviewer-prompt.md uses placeholders like {DIFF_OR_FILES} to inject external data into the prompt without explicit structural delimiters or instructions to ignore commands embedded within that data.
Capability inventory: The skill executes shell commands (git diff) and spawns multiple subagents. The lead reviewer agent is responsible for synthesizing findings and is explicitly instructed not to auto-apply changes.
Sanitization: There is no evidence of sanitization or filtering of the ingested content before it is processed by the subagent models.
[COMMAND_EXECUTION]: The skill uses local shell commands to determine the scope of the review.
Evidence: SKILL.md instructs the agent to run git diff main...HEAD to gather the changeset. This is a standard operation for development-oriented tools.

interrogate