qa-level-assessment

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill possesses an attack surface for indirect prompt injection because it ingests and processes untrusted external data in the form of PR descriptions and code diffs.
  • Ingestion points: SKILL.md (Phase 1) explicitly requests PR URLs or diff content from the user.
  • Boundary markers: None. The instructions do not define delimiters or specific safety warnings to ignore embedded instructions within the ingested PR data.
  • Capability inventory: None. The skill does not contain executable scripts, subprocess calls, network operations, or file-writing capabilities. It only generates textual assessment results.
  • Sanitization: None.
  • [Overall Assessment] (SAFE): No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill is entirely descriptive and instruction-based.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:48 PM