The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an evaluation loop that processes responses from an external AI Gateway, exposing the analyzing agent to potential indirect prompt injection (Category 8).
Ingestion points: Untrusted data enters the agent context via network responses in runner.ts (stored in JSON files) and is subsequently loaded for analysis as described in SKILL.md.
Boundary markers: The skill does not employ specific delimiters or isolation markers when presenting agent outputs within its JSON and Markdown reports, making it difficult for a judging AI to distinguish between instructions and data.
Capability inventory: The skill environment possesses capabilities such as shell command execution (git via Bun.spawn in runner.ts), network communication (fetch in runner.ts), and file system writes (writeFile in runner.ts and reporter.ts).
Sanitization: There is no evidence of sanitization or content filtering applied to the responses fetched from the AI Gateway before they are interpreted by the agent or included in the final markdown reports.

eval