agent-ai-ml-ops-specialist

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill configuration defines a legitimate specialist persona without introducing malicious logic or bypass instructions.
  • [COMMAND_EXECUTION]: The skill provides illustrative bash commands for ML tools (MLflow, Feast, BentoML) intended for reference; no arbitrary or dangerous command execution was identified.
  • [PROMPT_INJECTION]: The skill has a structural vulnerability to indirect prompt injection due to its operational design. Evidence: (1) Ingestion points: The agent uses WebSearch and WebFetch for external data and is instructed to read ~/.claude/skills/ai-ml-ops/SKILL.md. (2) Boundary markers: No explicit delimiters are present in the instructions. (3) Capability inventory: The agent has access to Bash, Write, and Edit tools. (4) Sanitization: No content validation is specified. This risk is inherent to the agent's research-oriented role and is assessed as safe in this context.
  • [EXTERNAL_DOWNLOADS]: The agent is equipped with standard web-fetching capabilities used for information retrieval; no automated remote script execution patterns were found.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 01:13 AM