oneshot-websites
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts (scripts/prepare_run.py, scripts/build_catalog_index.py, scripts/validate_catalog.py) to manage experiment metadata and index artifacts. These scripts are executed via the command line to maintain the environment and provide validation services.
- [PROMPT_INJECTION]: The skill generates artifacts based on user-supplied input, creating a surface for indirect prompt injection. This is addressed through robust defensive instructions in SKILL.md and the agent role definitions. Ingestion points: artifact/PROMPT.md. Boundary markers: Extensive instructions in SKILL.md and agents/oneshot-lead.md explicitly direct subagents to ignore embedded operational instructions. Capability inventory: Subagent spawning, file system writes, and tool access. Sanitization: Implemented via scripts/runtime_contract.py which enforces bounded reads and strict JSON parsing limits.
Audit Metadata