GitHub Prior Art Research
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [Indirect Prompt Injection] (SAFE): The skill instructs the agent to ingest and process untrusted content from GitHub searches. 1. Ingestion points: External data retrieved via WebSearch from github.com (READMEs, issues, and discussions). 2. Boundary markers: Absent; instructions do not provide delimiters for searched content. 3. Capability inventory: WebSearch; no dangerous file system or network operations are requested. 4. Sanitization: Absent. This behavior is standard for a research-oriented skill.
- [Prompt Injection] (SAFE): No malicious override patterns or safety filter bypass instructions were found. The instructional language used is benign.
Audit Metadata