The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from the local project, including source code, configuration files, and git history. This creates a surface for indirect prompt injection where malicious instructions embedded in the reviewed files could attempt to influence the agent's behavior during the audit.
Ingestion points: Local source files (src/pages/**, src/config.ts), git log output, and build/test results.
Boundary markers: The instructions do not define specific delimiters to separate the content being reviewed from the agent's internal instructions.
Capability inventory: The skill specifies the execution of npm run build, npm test, astro preview, and git log commands.
Sanitization: No explicit sanitization or filtering of the ingested project content is described.
[COMMAND_EXECUTION]: The instructions direct the agent to execute shell commands such as npm run build, npm test, and git log to verify project correctness and consistency. While these are standard development operations, they are performed within a workflow that processes potentially untrusted local code.

website-review