agent-research
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large amounts of untrusted data from the internet via
WebSearchandWebFetchtools. This content is then used to generate a final research report. The instructions lack explicit defensive measures such as boundary markers or sanitization to prevent the agent from executing instructions hidden within external web content. - Ingestion points: Data from external websites retrieved during the research phases in
SKILL.mdand processed via the templates inreferences/agent-prompts.md. - Boundary markers: Absent; there are no specific instructions to the specialist agents to differentiate between research data and potential embedded commands.
- Capability inventory: The skill possesses significant capabilities including creating new teams (
TeamCreate), spawning sub-agents (Agent), registering tasks (TaskCreate), and inter-agent communication (SendMessage). - Sanitization: Absent; the gathered text is processed analytically without filtering for malicious prompt patterns.
Audit Metadata