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