The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as git rev-parse, git log, and git diff. These commands use placeholders like {BASE_SHA} and {HEAD_SHA} that are populated at runtime. If these variables are sourced from untrusted inputs (e.g., external pull request data), it creates a potential surface for command injection.
[PROMPT_INJECTION]: The code-reviewer.md file serves as a prompt template that interpolates external data (such as {WHAT_WAS_IMPLEMENTED} and {PLAN_OR_REQUIREMENTS}) into the instructions for a subagent. This identifies an indirect prompt injection surface where malicious content in the requirements or implementation description could influence the reviewer subagent's behavior.
Ingestion points: Inputs are accepted into the code-reviewer.md template via several placeholders including {DESCRIPTION} and {PLAN_REFERENCE}.
Boundary markers: The template does not utilize delimiters or specific instructions to isolate the untrusted content from the system instructions.
Capability inventory: The agent has the ability to execute shell commands (git diff) and initiate tasks via the subagent superpowers:code-reviewer.
Sanitization: There is no evidence of input validation or sanitization for the interpolated strings.

requesting-code-review