The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions involve using the hermes CLI to list and configure agent profiles (hermes profile list, hermes -p <profile> model) via the agent's terminal tools.
[DYNAMIC_EXECUTION]: The skill manages agent lifecycles by dynamically creating, linking, and monitoring tasks through the kanban_create, kanban_link, and kanban_complete tools. It also leverages a 'goal mode' which utilizes an auxiliary judge agent for iterative verification of task completion.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a router that ingests user prompts and generates instructions for downstream worker agents. This creates a surface where malicious instructions in a user request could be propagated to other agents.
Ingestion points: User-supplied goals and multi-lane task descriptions processed in the decomposition playbook.
Boundary markers: Absent; the skill encourages sketching graphs and clarifying intent but does not specify technical delimiters for isolating task content.
Capability inventory: Access to kanban_create, kanban_link, kanban_complete, kanban_block, and kanban_reclaim tools which influence other agent processes.
Sanitization: Absent; user intent is decomposed and interpolated directly into new task bodies.

kanban-orchestrator