The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional Markdown and reference documentation without any executable scripts or external code dependencies.
[SAFE]: Includes explicit instructions for the agent to audit code for security issues, such as injection vulnerabilities and hardcoded secrets, as part of the review process.
[COMMAND_EXECUTION]: Utilizes standard git diff commands to establish the context of code changes, which is a legitimate and expected use of development tools for a code review skill.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted code inputs.
Ingestion points: Reads code targets (files, PRs, and diffs) as defined in SKILL.md.
Boundary markers: None present to explicitly separate reviewed code from agent instructions.
Capability inventory: Limited to read-only git operations for fetching diffs.
Sanitization: None present.
Assessment: The risk is negligible because the skill's primary purpose is critical analysis and reporting rather than execution of the ingested data.

code-review