The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several system commands to perform its core functionality. In SKILL.md (Step 1), it uses git status, git log, and git diff to establish the change baseline. In Step 4 (Fix Loop), it executes test runners such as pytest and various frontend testing tools to verify the correctness of applied fixes.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its handling of untrusted external content.
Ingestion points: Code diffs and PR metadata (description, comments) are fetched and processed in SKILL.md (Step 1).
Boundary markers: The instructions do not define clear delimiters or "ignore instructions" directives to prevent the agent from following malicious commands embedded within the reviewed code.
Capability inventory: The agent has the ability to execute shell commands (git, pytest) and write to the filesystem during the automated Fix Loop (Step 4).
Sanitization: No explicit sanitization or validation of the ingested data is described before it is integrated into the agent's context.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic execution patterns by modifying project source code and subsequently executing the resulting files via test runners in Step 4. It also employs a subagent pattern for adversarial re-reviews, creating new agent instances at runtime to validate code changes.

code-review