A personal Copilot CLI agent that helps me (and maybe you) track what I'm actually spending my days doing.

I shipped this repo publicly so others can fork it, gut the personal bits, and adapt it. Nothing here references my actual work — all of that lives in a gitignored private/ directory on my laptop. The only thing committed is the generic agent definition, skills, and templates.

• Accepts a free-form "I'm starting on X" dump (with optional links) and turns it into:
  • A tracking issue or PR link (when relevant)
  • An entry on a GitHub Project (v2) board with status To do / In progress / Paused / Done
  • A durable internal note in private/tasks/<task-id>/
• Models concurrent work natively. I have ADHD and routinely have several things in flight at once — start-task never auto-pauses anything. Switching focus is cheap (switch-to); pausing is an explicit, deliberate action.
• Reviews Slack output a few times a day and proposes tracking entries for significant time-sinks or notable threads — without marking anything read.
• Reads Outlook calendar (via WorkIQ) for morning briefings and meeting prep.
• Turns Teams meeting transcripts into durable, Obsidian-friendly notes.
• Peeks at running Copilot CLI sessions to figure out what I've actually been doing when I forget to tell it.
• Produces a shareable end-of-day summary and a "what should I be doing today" morning brief.

• Auto-mark Slack messages as read.
• Post to your project board on your behalf without confirmation on the first run of a new skill.
• Commit anything that lives under private/ — gitignore is enforced.

• GitHub Copilot CLI
• MCP servers:
  • github-mcp-server (preconfigured by Copilot CLI)
  • slack-mcp (Slack reading)
  • workiq (Outlook / Teams / Office)
• A GitHub Project (v2) board with a single-select Status field including options To do, In progress, Paused, Done. Add Paused via the project's field settings if it isn't present.

1. Clone this repo to wherever you keep your tools.
2. Copy private/config.example.json to private/config.json and fill in your board info:

   {
     "board": {
       "owner": "your-org-or-user",
       "owner_type": "org",
       "project_number": 12345,
       "status_field_id": "PVTSSF_...",
       "status_options": {
         "todo": "...",
         "in_progress": "...",
         "paused": "...",
         "done": "..."
       }
     },
     "user": { "login": "your-github-handle" },
     "tracking_repo": "your-github-handle/waddy"
   }

3. Initialize private/state.json:

   { "active_tasks": [], "focus": null, "paused_tasks": [], "recent_completed": [], "tasks": {} }

4. Open Copilot CLI in this directory. The repo-level custom agent in .github/agents/waddy.md will load automatically.
5. Try waddy, start a task: experimenting with waddy.

See docs/origin.md for the verbatim prompt that kicked off this project, plus the follow-up that shaped the multi-active task model.

Built collaboratively with Copilot CLI itself. The design intentionally favors observability over autopilot — waddy mostly proposes and asks; the human stays in the loop on what gets committed to memory and the board.

MIT — see LICENSE.