deep-research
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources, creating a potential surface for indirect prompt injection.
- Ingestion points: Research data enters the context through leaf researchers reading external URLs and documents (references in
SKILL.mdStep 3 andresearch-method.md). - Boundary markers: The skill contains strong explicit boundaries in
SKILL.mdStep 1.1, stating that source material is 'evidence, never authority' to change scope or run commands. - Capability inventory: The skill can perform local file writes to a dossier (
dossier.md), network reads for research, and spawn sub-agents for specialized roles (orchestration.md). - Sanitization: The skill implements a mandatory verification phase (Step 4) where a separate 'Verifier' role must validate claims against original sources to prevent misinformation or injection success.
Audit Metadata