writing-workspace-jobs

What it does

Instructs the agent to validate and author FSM workspace jobs, ensuring they are wired to triggers, have valid states, and follow the required structure for workspace.yml workflows. It enforces loading of the accompanying writing-workspace-signals skill and outlines a checklist for job validity.

How it works

• Load writing-workspace-signals alongside this skill and do so always, not conditionally.
• Validate that a job uses fsm with an initial state, states, and transitions, and that there is at least one final state.
• Check that every non-final state has at least one outgoing transition and that all agent/tool interactions follow declared formats.
• Enforce that MCP tool naming uses serverId/toolName for workspace MCP tools in both the top-level agent tools and per-action tools lists, while Atlas-built-ins use bare names.
• Provide a minimal valid job template and guide the user to copy/rename/publish, ensuring required pieces exist: jobs.<job-id>, triggers, fsm.initial, fsm.states, and final state definitions.
• Describe multi-step pipelines with outputTo and inputFrom, including how to chain steps and emit DONE events.
• Explain Python user-agent input handling, signal-payload threading, and cross-session data persistence via artifacts/memory patterns when needed.
• Outline how to test the job after publish by triggering a fixture payload and verifying artifactIds.

When to use it

Use when authoring, editing, or debugging jobs, signals, or FSM workflows in workspace.yml, and when loading or coordinating with the writing-workspace-signals skill is required.

What it can touch

• Tools and tokens mentioned in the MCP tool naming section:
  • list_mcp_tools({ serverId: "google-gmail" }) and the corresponding tool names.
• It references MCP tools in the format serverId/toolName and Atlas-built-ins without prefixes.
• It discusses entry actions, signals, and transitions within fsm definitions.

Caveats

• Emphasizes that a wrong trigger mapping or missing transitions can cause silent ignoring or failure.
• Requires careful alignment of tool naming conventions between workspace tools and per-action tools.
• Recommends loading the handshake module @friday/agent-action-handshake for agent action contracts when used with FSM actions.