limitr-py
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the application context which is then evaluated by the Limitr engine's policy logic.
- Ingestion points: Data enters the agent context via the
context,metadata, andeventarguments in methods includingallow,check,reserve,settle, andfeatures, as well as through themetadatafield inensure_customer. - Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are provided for the data passed into these methods to prevent adversarial context from influencing the policy engine's decisions.
- Capability inventory: The engine evaluates Stof policies, which can include logic for gating features. It can also execute registered Python host functions via
limitr.doc.liband perform HTTP requests if explicitly enabled vialimitr.doc.allow_http_hosts. - Sanitization: The documentation does not mention input sanitization or schema validation before the data is processed by the policy logic.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
limitrlibrary and its dependencies (stof,websockets) from the Python Package Index. - [COMMAND_EXECUTION]: Instructions provide shell commands for environment setup, specifically the installation of the required library.
- [DYNAMIC_EXECUTION]: The skill facilitates the dynamic loading and execution of Stof-based policies.
- Evidence: The
Limitr()constructor accepts and compiles policy strings (Stof, JSON, YAML, or TOML), and thedoc_callmethod allows for the runtime execution of functions defined within these policies.
Audit Metadata