About this skill

Problem to Solve

ThinkForce is a multi-agent orchestration platform. When an AI agent calls its REST API directly, it can get stuck discovering companyId, assigning agents after task decomposition, deciding which dependent steps are runnable, and treating done as a clean success.

This skill turns those platform rules into operational conventions, helping the model understand the state machine and execution boundaries before calling the Missions API.

How the Skill Works

Bootstrap: Call GET /api/companies first, then resolve and cache companyId instead of asking the user.
Task model: A Mission is a project and a Subtask is a step; subtasks form a DAG through dependsOn[] and nextSubtaskIds[].
Agent assignment: Decomposition usually does not assign agents; choose by agentName, agentRole, and enabledTools[], then PATCH assignedAgentId.
Execution control: Run only steps whose dependencies are complete, and check completedWithErrors to avoid ignoring tool errors.

Boundaries and Caveats

A valid X-TF-API-Key is required, and the key is bound to one company; switching companyId mid-session is not supported.
Per-step temporary tool attachment is not supported; choose an agent that already has the needed enabledTools[].
needs_attention requires human input or an explicit user instruction and does not continue automatically.

Screenshots

Use Cases

Break a market-research goal into multiple subtasks, assign agents by name and enabled tools, and track execution with status checks.
Create a ThinkForce mission, decompose it automatically, patch assignedAgentId for each subtask, and run only assigned steps using the Missions API.
Inspect blocked_upstream, pendingDeps, and completedWithErrors to decide whether to fix an upstream step or retry a dirty done subtask in the workflow.
Update runInstructions for a follow-up requirement, reset state, and re-run a paused or failed subtask in a live mission before shipping.

Best For

Agent engineers integrating ThinkForce who need to call the Missions API according to its state machine.
Platform integration engineers assigning appropriate agents to subtasks in automated workflows.
Application developers tracking DAG dependencies, failed retries, and tool errors in subtask output.
Automation owners creating, pausing, and resuming missions through the ThinkForce REST API.