The Agent Skills Directory

Prompt Injection (LOW): This skill is vulnerable to indirect prompt injection (Category 8) because its core function is to process untrusted external data.
Ingestion points: The agent uses Read, Grep, and WebFetch to ingest paper content provided by the user or fetched from URLs.
Boundary markers: The instructions do not define explicit boundary markers or 'ignore' instructions for the content being audited.
Capability inventory: The skill utilizes Read, Grep, Glob, WebSearch, and WebFetch. This combination allows the agent to read local files (like ./docs/domain-model.md) and make external network requests.
Sanitization: There is no evidence of sanitization or filtering of the ingested content before processing.
External Downloads (SAFE): The WebFetch and WebSearch tools are used specifically for the primary purpose of retrieving academic papers for audit, which is consistent with the skill's stated goal.

argument-audit