deep-architecture-research

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust 'confirmation gate' strategy, requiring the agent to present a research map (framing brief) and obtain user consent before performing any substantive research tasks. This prevents unauthorized tool usage or unintended data collection.
  • [SAFE]: The 'Evidence Protocol' promotes high-integrity research by prioritizing actual source code and official documentation over secondary community signals. It explicitly instructs the agent to separate verified facts from historical data or roadmap plans.
  • [SAFE]: The skill encourages the use of local clones for source code analysis rather than browsing individual web pages, which is a best practice for understanding complex system architectures and call paths without triggering excessive external requests.
  • [SAFE]: Data handling is restricted to creating durable Markdown artifacts within the project's research location, with instructions to keep scratch notes out of shared paths, supporting good data hygiene.
  • [SAFE]: No evidence of prompt injection, obfuscation, or unauthorized network activity was found. The instructions focus entirely on improving the quality and traceability of technical research.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 06:13 PM
Security Audit — agent-trust-hub — deep-architecture-research