The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted code changes retrieved via version control commands. Malicious instructions embedded in a code commit could potentially influence the agent's behavior during the review process.\n
Ingestion points: Untrusted code diffs and file content retrieved using git diff (referenced in SKILL.md).\n
Boundary markers: The skill does not implement delimiters or instructions to isolate the code content from the agent's operational logic.\n
Capability inventory: The skill uses local CLI tools such as git, npm, pnpm, and npx to perform its tasks (referenced in SKILL.md and checks/dependencies.md).\n
Sanitization: There is no evidence of sanitization or escaping of the input diff data.\n- [COMMAND_EXECUTION]: The skill leverages standard development and auditing tools to perform its primary function of reviewing code.\n
Evidence: Integration with git for diffing, npm/pnpm for dependency auditing, and npx for running analysis tools (found in SKILL.md, checks/dependencies.md, and checks/git.md).\n- [EXTERNAL_DOWNLOADS]: The skill utilizes npx to run established community tools for dependency and license auditing from the NPM registry.\n
Evidence: References to npx depcheck and npx license-checker in the checks/dependencies.md file.

review