fabled
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill performs an external network request to
https://www.demandsphere.com/research/demandsphere-radar/ai-frontier-model-tracker/api.jsonusingcurlto fetch model performance rankings. This domain is not a recognized trusted organization or well-known technology service. - [COMMAND_EXECUTION]: The skill instructs the agent to execute a provided Python script (
scripts/check.py) to validate output. It also directs the agent to write and execute unit tests against the code generated in response to user prompts (Phase 4, Mode B). While intended for automated verification, this entails executing code derived from potentially untrusted input. - [PROMPT_INJECTION]: The skill architecture presents an indirect prompt injection surface because it takes untrusted user prompts and uses them to generate and execute code without explicit sanitization.
- Ingestion points: User-supplied build prompts provided as input to the agent.
- Boundary markers: None used to encapsulate or delimit user-supplied content within the internal generation process.
- Capability inventory: Network access (curl), file system access, and command execution (via subprocess for scripts and generated tests).
- Sanitization: No validation or sanitization is performed on the user prompt before it influences logic generation and execution.
Audit Metadata