failure-to-guardrail
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a local Python script (
scripts/validate-guardrail-record.py) to validate guardrail records. The script is constrained to reading a JSON file provided as an argument and validating its fields against a fixed schema using the Python standard library. It does not invoke sub-processes or external shells. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data (AI failure logs) to create reproduction cases. However, the ingestion is structured via a JSON template (
assets/guardrail-record.template.json), and the validation script treats the input purely as data. No execution path exists for instructions embedded within the failure records. - [DATA_EXPOSURE]: No hardcoded credentials, sensitive file path access, or network exfiltration patterns were detected. The script's file access is limited to the path provided as a command-line argument for validation purposes.
- [EXTERNAL_DOWNLOADS]: The skill does not define any external dependencies or remote code execution patterns. It operates entirely on local files provided within the skill package.
Audit Metadata