worker-benchmarks
Warn
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTIONNO_CODE
Full Analysis
- [EXTERNAL_DOWNLOADS] (MEDIUM): The documentation recommends running 'npx agentic-flow'. The package 'agentic-flow' is not among the verified trusted sources, meaning the agent would download and execute unvetted code from the npm registry.
- [PROMPT_INJECTION] (MEDIUM): The skill defines benchmarks for 'trigger-detection' and 'worker registry' operations. (1) Ingestion points: Worker triggers and registry metadata are processed during benchmarking. (2) Boundary markers: No markers are defined to separate untrusted trigger data from the benchmark instructions. (3) Capability inventory: The benchmark results influence agent selection and configuration settings. (4) Sanitization: No sanitization is mentioned for the triggers being evaluated.
- [NO_CODE] (LOW): This skill provides no local scripts or binaries; it consists entirely of markdown instructions that direct the agent to perform external operations.
Audit Metadata