The Agent Skills Directory

[SAFE]: The skill's primary function is technical code review, which is a legitimate development workflow. It focuses on validating implementation consistency against upstream documents.
[COMMAND_EXECUTION]: The skill instructs the agent to use git diff for comparing branches or commit ranges. This is a standard local repository operation and does not involve executing arbitrary shell commands from untrusted inputs.
[DATA_EXFILTRATION]: The skill does not perform any network operations or external data transmissions. Its activities are confined to reading local project files and updating a pipeline.md file to track review status.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes external data such as code comments and issue descriptions.
Ingestion points: Documentation files (prd.md, spec.md, tdd.md), issue tracker files, and source code being reviewed.
Boundary markers: Absent; the instructions do not define specific delimiters for separating untrusted code content from the agent's instructions.
Capability inventory: File system read/write (updating pipeline.md) and git diff execution.
Sanitization: None identified; the skill relies on the LLM's internal safety guardrails to differentiate between implementation code and operational instructions.

review