The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions for task orchestration and management. It emphasizes the principle of 'route, don't execute', which acts as a safety constraint by preventing the orchestrator agent from performing potentially high-risk implementation tasks directly. The use of environment variables for tenant identification is a standard practice for multi-user session management and does not expose sensitive credentials.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting user goals and decomposing them into tasks for other specialist agents, which creates a potential surface for indirect prompt injection. 1. Ingestion points: User-provided project goals and clarifying questions (Step 1). 2. Boundary markers: The skill explicitly includes a human-in-the-loop boundary (Step 2), requiring the agent to show the proposed task graph to the user for review before creating tasks. 3. Capability inventory: The skill uses task management tools (kanban_create, kanban_complete, kanban_link, kanban_block) to route work but does not include tools for direct code execution, file system modification, or network access. 4. Sanitization: No specific sanitization or escaping of user input is described, but the required human review of the task graph serves as a significant mitigation.

kanban-orchestrator