The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted source code and git diffs to be processed by specialized sub-agents. Malicious instructions embedded in the code being reviewed (e.g., in comments) could potentially subvert the reviewer agents' logic.\n
Ingestion points: SKILL.md (Workflow Step 1) retrieves file contents and git diffs based on the $ARGUMENTS provided.\n
Boundary markers: Absent. The instructions do not specify any delimiters or warnings to the sub-agents to ignore instructions contained within the code context.\n
Capability inventory: Bash, Read, Glob, Grep, TeamCreate, and SendMessage are available to the agent and its sub-agents.\n
Sanitization: Absent. The skill does not perform any pre-processing or sanitization on the retrieved code before passing it to reviewers.\n- [COMMAND_EXECUTION]: The workflow in SKILL.md interpolates the $ARGUMENTS variable directly into shell commands such as gh pr diff $target and git diff main...$target. If the environment does not provide strict argument sanitization, this pattern could allow for command injection through malicious input strings containing shell metacharacters.

review