The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses the bunx utility to download and execute the @willbooster/agent-skills package from the npm registry at runtime.
[COMMAND_EXECUTION]: Executes shell commands via the Bash tool to perform the review. The instructions include a specific directive to not stop the command for at least one hour, which is an unusual requirement that forces the agent to keep a process active for an extended period.
[EXTERNAL_DOWNLOADS]: Fetches the @willbooster/agent-skills package from the npm registry using the @latest tag, which lacks version pinning and always pulls the most recent code.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface because it processes untrusted content from the codebase being reviewed. 1. Ingestion points: Code files within the current branch. 2. Boundary markers: Absent; there are no instructions to the agent to differentiate between the code content and instructions. 3. Capability inventory: The Bash tool is available to the agent for executing commands. 4. Sanitization: Absent; no sanitization or content validation is performed on the files before they are processed by the tool.

review-gemini