qrspi-create-research
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from a user-provided research questions document, which represents a potential surface for indirect prompt injection. However, the risk is mitigated by explicit constraints directing the agent to only document current reality and specifically forbidding the suggestion of changes or architectural critiques.
- Ingestion points: Reads content from a user-provided path, typically
thoughts/shared/questions/YYYY-MM-DD-ENG-XXXX-questions.md. - Boundary markers: No specific delimiters or "ignore instructions" warnings are applied to the imported questions text.
- Capability inventory: Spawns sub-agents (
codebase-locator,codebase-analyzer,codebase-pattern-finder,thoughts-locator), reads codebase files using the Read tool, and writes research reports tothoughts/shared/research/. - Sanitization: The skill does not implement automated sanitization or filtering of the input questions file content, relying instead on high-level instructional constraints.
- [DATA_EXPOSURE]: The skill is designed to read and analyze codebase files and internal documentation. This access is limited to the local project environment for the purpose of generating research reports and does not involve any network exfiltration or access to sensitive credentials like SSH keys or environment secrets.
Audit Metadata