Your first skill

After this page you will have run one NeuroDock skill end-to-end and seen the substrate work as a chain: chronometric → cognitive-graph → adhd-daily-planner.

sequenceDiagram
  autonumber
  participant You
  participant Claude
  participant Chrono as mcp-chronometric
  participant Graph as mcp-cognitive-graph
  participant Skill as adhd-daily-planner

  You->>Claude: /morning-brief
  Claude->>Chrono: get_time_context()
  Chrono-->>Claude: { local_time, day, energy_zone }
  Claude->>Graph: recall_decisions(project, since)
  Graph-->>Claude: [] (empty on first run)
  Claude->>Skill: assemble brief
  Skill-->>Claude: 1-paragraph brief + 1-3 priorities
  Claude-->>You: morning brief

What /morning-brief does under the hood. A short screen-capture demo is tracked in issue #27.

Prerequisites

• Installation is complete and neurodock doctor passes.
• Claude Desktop is restarted (full quit, not close-window) and you can see the five NeuroDock MCP servers in its tool list.

Run it

Open Claude Desktop, start a new chat, and type /morning-brief. That’s it.

You should see:

• A tool call to get_time_context() from mcp-chronometric returning the current time, your energy zone, and how long since your last prompt.
• A tool call to recall_decisions(project: ..., since: ...) from mcp-cognitive-graph returning any decisions made overnight (empty on a fresh install).
• A one-paragraph brief plus 1–3 things that matter today, assembled by the adhd-daily-planner skill.

Expect the first run to be sparse

On a fresh install the cognitive graph is empty, and the brief will say so plainly — something like:

The cognitive graph has no decisions or open blockers recorded yet. That is expected on a fresh install. Today’s brief is just the time context: morning peak, Friday, no open session.

This is correct behaviour. The substrate does not fabricate context it does not have. As you use NeuroDock — recording decisions, naming projects, closing sessions — the brief gets richer.

What happened

The skill called two MCP tools (no LLM call inside the substrate) and assembled the result. None of this leaves your machine: the profile, the graph, the session state — all local.

What’s next

• Edit your profile so the substrate matches your rhythms.
• Look at the adhd-daily-planner source in packages/skills/adhd-daily-planner/SKILL.md to see exactly what the skill does. Skills are markdown bundles, not opaque code.
• Try /resume (the audhd-context-recovery skill) for the inverse experience: yesterday’s mental state, reconstructed from the graph.