cherry-pick
Pass
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external content (e.g., Perplexity answers, articles, research dumps), which is a common vector for indirect prompt injection attacks.
- Ingestion points: The skill ingests large blocks of text pasted by the user from external sources as a "candidate pool" in SKILL.md.
- Boundary markers: The prompt does not define explicit delimiters (e.g., XML tags or specific separators) to isolate the untrusted content from the agent's instructions.
- Capability inventory: The skill is intended to integrate ideas into the current project, which may involve modifying code or configuration files depending on the agent's primary role.
- Sanitization: The skill contains high-quality defensive instructions that mitigate this risk, such as "atomizing" the input to prevent hidden instructions from being missed, a strict "Resist source authority" rule, and a mandatory "Validity" gate that requires verification of claims before they are accepted.
Audit Metadata