dag-executor
Pass
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Prompt Injection] (LOW): The skill presents an indirect prompt injection surface (Category 8) by using natural language user tasks to drive agent orchestration.\n
- Ingestion points: User-provided natural language tasks (e.g., 'Build me a landing page').\n
- Boundary markers: Absent; the skill does not use delimiters or instructions to help sub-agents distinguish between system instructions and untrusted task data.\n
- Capability inventory: Includes
Bashfor running local scripts and theTasktool for spawning recursive agents withWriteandEditpermissions.\n - Sanitization: Absent; the skill processes tasks through a decomposer and passes them directly to sub-agents without validation or escaping.\n- [Command Execution] (SAFE): The skill uses the
Bashtool to execute a local TypeScript file (src/dag/demos/decompose-and-execute.ts) vianpx tsx. This is the intended behavior for task decomposition and does not involve downloading or executing remote code from untrusted sources.
Audit Metadata