strata

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data such as live market capabilities, Sonar quotes, and maker reputation from the Strata MCP server (https://api.stratabook.app/mcp).
  • Ingestion points: Market listing data, Sonar quote responses, and market-making status updates from the external API are read into the agent's context in SKILL.md and references/public-contract.md.
  • Boundary markers: The instructions include boundary checks, requiring the agent to "Base explanations on the returned quote fields; do not invent a breakdown that is not present" and to verify all economically meaningful inputs.
  • Capability inventory: The skill provides tools for preparing and submitting transactions (e.g., strata_market_making_prepare, strata_market_making_submit_and_wait), as detailed in the action graph.
  • Sanitization: The instructions mandate preserving atomic amounts as base-10 strings and emphasize that all signing must remain external to the platform.
  • [COMMAND_EXECUTION]: The documentation in SKILL.md and references/public-contract.md describes using npx -y @stratabook/sdk for terminal discovery, scripted action primitives, and non-trading latency certification (order-slo).
  • [EXTERNAL_DOWNLOADS]: The skill relies on external vendor resources, including the @stratabook/sdk Node package and the official Strata MCP server hosted at https://api.stratabook.app/mcp.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 03:39 AM