gtd-deliberation
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill relies on external research and web searches to verify data freshness, which creates a vulnerability surface for malicious instructions embedded in third-party content.
- Ingestion points: External research findings, authoritative sources, and web search results ingested during the 'Freshness Search' phase described in
references/freshness-search.md. - Boundary markers: The skill includes a dedicated
references/security-boundaries.mdfile that explicitly commands the agent to ignore instructions embedded in retrieved content and to treat such content as evidence rather than instructions. - Capability inventory: The skill includes a Python script (
scripts/deliberation.py) for file system operations (creating and validating JSON models) and utilizes network search capabilities provided by the agent environment. - Sanitization: Instructions require the agent to reduce search queries to public facts and to surface conflicting evidence rather than following instructions from external sources.
Audit Metadata