optimization-theory-expert
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The research checklist in modules/research-checklist.md directs the agent to fetch external data to update the skill's instructions. This ingestion of untrusted data from sources like Wikipedia or Arxiv, combined with the capability to modify the skill's core guidance, creates an attack surface for indirect prompt injection.
- Ingestion points: External sources retrieved via tool calls in modules/research-checklist.md.
- Boundary markers: The skill lacks explicit delimiters or warnings to ignore embedded instructions within the fetched data.
- Capability inventory: The agent is instructed to write and update instructions in modules/core-guidance.md and metadata.
- Sanitization: No sanitization or validation of the ingested external content is performed before interpolation.
Audit Metadata