reflect-appworld-failure

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external task descriptions and error logs. These inputs are interpolated into the prompt without explicit instructions to disregard potentially malicious commands embedded within the data, creating a surface for indirect prompt injection that could influence the generated playbook bullets.
  • Ingestion points: Input data sections including # Task, ## Error Messages, and ## Failed Code Snippet (SKILL.md).
  • Boundary markers: Uses markdown headers but lacks explicit 'ignore embedded instructions' delimiters for untrusted sections.
  • Capability inventory: The skill is authorized to use the Read tool.
  • Sanitization: No sanitization or escaping of external content is implemented before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:30 PM