error-recovery

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill includes instructions to 'Rephrase prompt -> Retry without triggering content filter' when encountering safety blocks, which is an attempt to circumvent guardrails.
  • [COMMAND_EXECUTION]: The skill performs automatic shell command execution, specifically 'mkdir -p', to resolve file system path issues.
  • [INDIRECT_PROMPT_INJECTION]: The skill's error handling loop ingests untrusted data. 1. Ingestion points: API provider and network error responses (SKILL.md). 2. Boundary markers: None provided to isolate external strings. 3. Capability inventory: file:read, file:write, and command execution. 4. Sanitization: No evidence of sanitization for error messages used to reformulate prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 07:41 PM
Security Audit — agent-trust-hub — error-recovery