nitro-task-orchestrator
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies extensively on the
nitroCLI tool for task tracking, agent communication, and shared memory management. It executes commands such asnitro agent tasks show,nitro agent mail inbox, andnitro agent memory contextto coordinate work across the multi-agent pipeline. - [INDIRECT_PROMPT_INJECTION]: The orchestrator and its subagents ingest untrusted data in the form of task descriptions and comments retrieved from the
nitrotracker. This represents an indirect prompt injection surface as malicious instructions could be embedded in the task backlog. - Ingestion points: Data enters the context via
nitro agent tasks show <id> --output jsonwithin theimplementPrompt,reviewPrompt, and other roles defined inSKILL.mdandassets/workflows/nitro-backlog-wave.js. - Boundary markers: The skill instructions emphasize role isolation and strict scope enforcement (e.g., "Implement exactly the ticket's scope").
- Capability inventory: The subagents have the capability to write files, perform git commits, and interact with the
nitroCLI infrastructure. - Sanitization: The workflow script enforces structured output using JSON schemas (
change-result.schema.json,review-result.schema.json, etc.) to validate subagent responses.
Audit Metadata