ic-feedback

Pass

Audited by Gen Agent Trust Hub on Jul 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill interacts exclusively with the official vendor domain (immersivecommons.com) for its API and installation resources.
  • [PROMPT_INJECTION]: The skill addresses potential indirect prompt injection from processing untrusted user feedback. Analysis of the attack surface:
  • Ingestion points: Data enters the context through feedback fields (message, expected, got, category) returned by the ic_admin_list_feedback tool in SKILL.md.
  • Boundary markers: The skill mandates using the message_safe_render field, which wraps untrusted content in a <USER_SUBMITTED_TEXT> envelope to prevent context escaping.
  • Capability inventory: The agent has access to administrative tools (ic_admin_resolve_feedback, ic_admin_approve_tier_request), but is strictly forbidden from invoking them based on input from the feedback queue.
  • Sanitization: The server performs Unicode normalization (NFKC) and strips control characters and zero-width characters prior to delivery.
  • [EXTERNAL_DOWNLOADS]: The skill provides installation instructions in INSTALL.md that download its configuration from the vendor's official domain.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 19, 2026, 10:21 PM
Security Audit — agent-trust-hub — ic-feedback