feedback-learning

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • Indirect Prompt Injection Surface: The skill facilitates the capture and persistence of user feedback into repository documentation. This introduces a surface where external input could potentially influence the agent in subsequent sessions if processed without isolation.
  • Ingestion points: User-provided corrections and design preferences are identified and processed for recording (Workflow Step 1).
  • Boundary markers: The documentation format does not include specific boundary markers or 'ignore' instructions for the persisted content (Learning inbox format).
  • Capability inventory: The skill allows the agent to read across project files and write updates to the .github/learnings/ directory and other design documents (Workflow Steps 2 and 5).
  • Sanitization: The instructions emphasize generalizing principles rather than incident chronology, but do not provide specific technical sanitization for user-provided data (Workflow Step 4).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 06:11 PM
Security Audit — agent-trust-hub — feedback-learning