The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted data from repository diffs and external documentation links.
Ingestion points: Code changes retrieved via git diff (SKILL.md) and external documentation content retrieved via WebFetch and WebSearch (references/contracts.md).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to isolate untrusted data from the system prompt.
Capability inventory: The agent has the ability to execute shell commands (git), perform network operations (WebFetch), and navigate the filesystem (ls, find).
Sanitization: No explicit sanitization or filtering of external content is described before it is provided to the agent for review.
[COMMAND_EXECUTION]: The skill executes the git diff command to retrieve code changes for analysis. This is a core component of its review functionality and is used in a read-only manner.
[EXTERNAL_DOWNLOADS]: The skill utilizes WebFetch and WebSearch to retrieve API documentation from the internet to verify contract compliance. These operations are performed dynamically based on the content of the code under review.
[SAFE]: The skill accesses tool-specific configuration files in the user's home directory (e.g., ~/.claude/CLAUDE.md, ~/.codex/AGENTS.md) to identify project-specific rules. These paths are specific to the operation of the supported AI agent platforms and do not involve sensitive credentials.

review-code