The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git diff using user-supplied strings for commit hashes and ranges in SKILL.md. While the variables are wrapped in double quotes, this pattern constitutes a potential command injection surface if the host platform does not rigorously sanitize user input.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted external data (source code and diffs) as part of its primary workflow. 1. Ingestion points: User-provided diffs and files read from the repository in Steps 1 and 2. 2. Boundary markers: None explicitly defined to isolate untrusted data from instructions. 3. Capability inventory: Subprocess execution via git diff and extensive file system read access. 4. Sanitization: The skill lacks explicit sanitization or escaping logic for the content it analyzes, relying on the agent's internal safety filters.

code-review