task-complexity-classifier
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted task descriptions provided by users or external input to determine task complexity through the TypeSafe (Jev) AI model.
- Ingestion points:
classifier.pyreads task descriptions directly from command-line arguments or standard input. - Boundary markers: No protective delimiters or explicit 'ignore embedded instructions' warnings are used when interpolating the untrusted task description into the classification request.
- Capability inventory: The skill performs network requests to the TypeSafe API. Its output is intended to influence downstream agent logic, such as selecting higher-tier AI models or allocating more resources for 'complex' tasks, which could be manipulated by an attacker providing a deceptive task description.
- Sanitization: No validation or sanitization of the input text is performed before it is sent to the external model.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python packages, specifically
typesafe-sdkandpython-dotenv, from standard package registries.
Audit Metadata