app-questions

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill collects untrusted user input through an interactive Q&A workflow and interpolates this data into CLAUDE.md, a file specifically designed to guide AI agent behavior. This creates a risk where malicious instructions provided by a user (or an external source influencing the user) could influence the agent's future actions. \n
  • Ingestion points: User answers during the 7-phase questionnaire and the interactive editor described in SKILL.md. \n
  • Boundary markers: The template for CLAUDE.md does not include delimiters or instructions for the agent to ignore potentially malicious content within the user-supplied fields. \n
  • Capability inventory: The skill is designed to write files (APP_DEFINITION.md and CLAUDE.md) and interact with the user. \n
  • Sanitization: There is no evidence of sanitization or validation of the user input before it is written to the persistent guidance files. \n- [CREDENTIALS_UNSAFE]: Phase 5 of the skill's discovery workflow explicitly asks the user for information regarding "secure secrets" and persists the response in plain text within APP_DEFINITION.md. This practice encourages users to store sensitive credentials in a project documentation file that might be shared or committed to version control, rather than using a dedicated secrets manager or environment variables.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 09:14 AM
Security Audit — agent-trust-hub — app-questions