The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, which could contain malicious instructions designed to influence the agent's behavior during triage or drafting.\n
Ingestion points: The agent processes external data via query_text and requirements fields provided from sources like HARO, Qwoted, and Featured as specified in the Inputs section.\n
Boundary markers: The skill instructions do not define specific delimiters or escaping mechanisms to separate the untrusted query content from the prompt context.\n
Capability inventory: The skill has no high-risk capabilities; it is restricted to text generation and explicitly prohibits automated sending of responses (e.g., 'Never auto-send').\n
Sanitization: No sanitization or filtering of the external query text is mentioned before the agent evaluates it for fit or drafts a response.

reactive-comment