axolotl

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill documents CLI commands for training, performance benchmarking (e.g., NCCL tests), and cloud interactions as seen in SKILL.md and api.md.
  • [EXTERNAL_DOWNLOADS]: Documentation includes functions like fetch_from_github and check_remote_config for downloading resources from GitHub and other HTTPS URLs.
  • [REMOTE_CODE_EXECUTION]: The documentation references the trust_remote_code setting for loading datasets, which is a standard but sensitive feature in the machine learning ecosystem.
  • [INDIRECT_PROMPT_INJECTION]: The skill operates as a guide for a framework that processes external datasets and YAML configurations.
  • Ingestion points: Input datasets (referenced in references/dataset-formats.md) and YAML configurations provided by the user.
  • Boundary markers: Relies on structured data formats (YAML/JSONL).
  • Capability inventory: CLI execution, file system manipulation, and network access (as documented in references/api.md and references/other.md).
  • Sanitization: Relies on the underlying Python libraries (e.g., PyYAML, datasets) and user discretion during dataset selection.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 07:01 PM
Security Audit — agent-trust-hub — axolotl