The Agent Skills Directory

[DATA_EXPOSURE]: The skill instructs the agent to read the global configuration file at ~/.claude/CLAUDE.md. While used for context, this path is outside the immediate project scope and may contain user-specific global standards.
[COMMAND_EXECUTION]: The skill executes various git commands and arbitrary shell commands provided by the user or project context (e.g., <test command>) to verify code issues. This allows for the execution of code present in the local repository.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted content from the codebase, such as new files and git diffs, which could contain malicious instructions designed to influence the agent's review output.
Ingestion points: Files identified via git status, git diff, and git ls-files are read in their entirety.
Boundary markers: There are no explicit delimiters or instructions to the agent to treat the code content as data only and ignore embedded instructions.
Capability inventory: The skill can execute shell commands (git, <test command>) and write files to the .agents/code-reviews/ directory.
Sanitization: The skill does not implement sanitization or escaping of the code content before analysis.

code-review