skills/dydoc/skills/to-questionnaire/Gen Agent Trust Hub

to-questionnaire

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text regarding recipient roles and specific information gaps and interpolates this data directly into a generated Markdown file (to-questionnaire-<slug>.md). Because the output file is intended to be read by others or potentially processed by other agents, the lack of sanitization or boundary markers creates a surface where malicious instructions could be embedded in the questionnaire text.
  • Ingestion points: User inputs defining the recipient's role, expertise, and specific information needs (SKILL.md, steps 1 and 2).
  • Boundary markers: The <questionnaire-template> lacks delimiters or explicit instructions to downstream processors to ignore embedded text.
  • Capability inventory: The skill utilizes file writing capabilities to create Markdown documents in the current directory (SKILL.md, step 3).
  • Sanitization: No sanitization or validation of the user-provided text is performed before it is written to the file.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 06:28 AM
Security Audit — agent-trust-hub — to-questionnaire