meta-methodology-research-methodology
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a meta-methodology for research (Glob -> Grep -> Read). It emphasizes 'Evidence First' principles, requiring actual code verification before any claims are made. The instructions explicitly prohibit writing or editing files, reinforcing a read-only posture that prevents unauthorized modifications to the codebase.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow for processing untrusted codebase content. However, the methodology focuses on structural discovery (pattern discovery, API cataloging) and requires evidence-based structured output (file paths and line numbers), which minimizes the risk of the agent executing embedded instructions from the analyzed files. The 'Self-Correction Triggers' further mitigate risks by instructing the agent to stop and verify if it begins making claims without direct evidence.
Audit Metadata