skills/fw-ai/cookbook/configure/Gen Agent Trust Hub

configure

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the vendor-provided firectl CLI for managing training jobs and a standard Python utility for generating session identifiers. These operations are within the expected scope of the tool.\n- [CREDENTIALS_UNSAFE]: Authentication is handled via secure terminal input (read -s), specifically instructing the agent never to echo or request secrets in the chat, which is a security best practice.\n- [EXTERNAL_DOWNLOADS]: The skill fetches live pricing and model configuration data from official Fireworks AI documentation. These are trusted vendor sources used for accurate cost estimation.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes dataset metadata (such as schema and row counts) to prepare training plans. \n
  • Ingestion points: Dataset schema and metadata evaluation in SKILL.md.\n
  • Boundary markers: The skill uses a structured 'Mandatory final-plan gate' to confirm all parameters with the user before execution.\n
  • Capability inventory: Resource creation and training management via firectl.\n
  • Sanitization: Input validation is performed locally and via the CLI before any remote operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 06:00 PM
Security Audit — agent-trust-hub — configure