create-research-questions
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from user-supplied files and external URLs, creating a surface for potential instruction injection.
- Ingestion points: The agent reads ticket files, task briefs, and external documentation via web search tools (Steps 1 and 2).
- Boundary markers: No specific delimiters are used to separate user data from instructions, though the skill explicitly scopes exploration to named files.
- Capability inventory: The skill can use repository analysis agents, perform web searches, and write markdown files to the local
.agent-work/tasks/directory. - Sanitization: The 'Current-state gate' in Step 5 requires the agent to audit its own output to ensure questions are descriptive and objective, which serves as a control against untrusted data influencing the artifact's logic.
Audit Metadata