limitr-ts

Pass

Audited by Gen Agent Trust Hub on Nov 21, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation guides the user to install the @formata/limitr and @formata/stof packages via npm. These packages are part of the vendor's ecosystem (dev-formata-io) and are required for the SDK's core functionality.
  • [COMMAND_EXECUTION]: The skill provides standard installation commands (npm i) to set up the necessary environment and dependencies.
  • [DYNAMIC_EXECUTION]: The skill utilizes the Stof engine, a WebAssembly-based policy runtime. This engine supports dynamic features such as registering host functions and executing Stof scripts (callSource), which are essential for its role as a flexible policy engine.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes integration recipes for governing AI agents (such as those using the Claude Agent SDK). These recipes pass decision information (e.g., "limit reached") from the Limitr engine back to the AI agent to inform its behavior.
  • Ingestion points: Usage of permissionDecisionReason in the governed function and return values in governedCreate provide feedback to the model.
  • Boundary markers: No specific delimiters for the decision messages are included in the provided code snippets.
  • Capability inventory: The SDK can execute dynamic logic and optionally perform network requests if configured by the user via the allowHttp method.
  • Sanitization: The why() utility function maps engine decision reasons to descriptive strings before they are sent to the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 21, 2026, 02:37 AM