ulw-research
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection due to its core function of processing untrusted data from the web and external repositories.\n
- Ingestion points: Data is retrieved via omh_search and by reading core modules of open-source reference implementations.\n
- Boundary markers: The skill instructions specify that external content should be treated as claims and distinct from agent instructions, with evidence clearly separated from inference.\n
- Capability inventory: Capabilities include network operations via omh_search and file system interactions for managing citation ledgers and research progress.\n
- Sanitization: Evidence is structured into a deep research dossier and validated through the omh.workflows.research_dossier module to ensure correct data schemas.\n- [SAFE]: Framework-specific tool calls and local script executions, such as omh_search and sources.py, align with the workflow's intended use and do not exhibit malicious behavior.
Audit Metadata