The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted code diffs from external sources such as pull requests and commits, which is a surface for indirect prompt injection where malicious instructions could be hidden in the code being reviewed. Ingestion points: The skill uses gh pr diff and git diff in SKILL.md. Boundary markers: The instructions lack specific delimiters for code content, though the critique step in critique_rules.md provides a logical filter. Capability inventory: Access to git, gh, and local python execution. Sanitization: Filename sanitization is performed in scripts/split_diff.py.
[COMMAND_EXECUTION]: The skill executes git and GitHub CLI commands, as well as a local Python script scripts/split_diff.py. These are appropriate tools for the skill's purpose and do not involve arbitrary command injection from user input.
[SAFE]: No obfuscation, hardcoded credentials, or unauthorized network activity was detected. The skill's implementation is transparent and follows its stated purpose.

code-review