The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it is designed to read and analyze arbitrary files within a repository during review tasks.
Ingestion points: Reads repository source files, policy definitions from context/config/review-policies/, and user-provided instructions via CLI arguments.
Boundary markers: The instructions lack explicit delimiters or safety prompts to ignore instructions embedded within the codebase being analyzed.
Capability inventory: The skill can execute shell commands (git), write files to the project directory (change.md, research/, reviews/), and invoke multiple other diagnostic and management skills.
Sanitization: There is no specified mechanism for sanitizing or escaping the content of files or policies before they are processed.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to inform its review process.
Evidence: Employs git log --oneline to identify frequently changed paths and git rev-parse HEAD to record the current commit.
Context: These are standard development operations, but they represent a functional capability that could be targeted if inputs were manipulated.

dx-review