internet
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements an investigation-based research protocol with extensive guidance on source verification and contradiction management. It includes clear instructions to exclude secrets and personal data from durable memory and mandates user approval for persistent writes.
- [INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data from the internet, which represents a potential indirect injection surface.
- Ingestion points: External web content, documentation, search results, and community platforms as specified in
SKILL.mdand thereferences/directory. - Boundary markers: The protocol requires the agent to clearly separate source-backed claims from its own inference and to record contradictions neutrally.
- Capability inventory: The skill utilizes a local Python script
scripts/research_workspace.pyto write logs, plans, and claim ledgers to a dedicated research directory on the filesystem. - Sanitization: The agent is instructed to perform source tier analysis and cross-verify claims; the supporting workspace script uses standard libraries to safely format data into CSV and Markdown files.
Audit Metadata