Bon — The Ticket
GTD-flavoured work tracking for Claude-human collaboration.
In a professional kitchen, the bon is the ticket — the slip of paper that tells the brigade what’s been ordered and what needs to go out. This bon does the same for knowledge work: it tracks what needs doing, in GTD vocabulary, so that both human and AI can pick up where they left off. No sprints, no story points, no velocity — just outcomes, actions, and “what can I work on now?”
When to Use / When NOT to Use
Use bon when:
- You need to track work across multiple sessions — outcomes that take days, actions that get interrupted
- You want a future Claude session to understand why something matters, not just what to do
- You’re breaking a large goal into concrete next steps
- You want to file work for later with enough context that a cold session can pick it up
Do NOT use bon when:
- The task fits inside a single session — just do it
- You need project management features (dependencies, timelines, resource allocation) — bon is deliberately not that
- You want a shared team tracker — bon is per-workspace, for one human-AI pair
- You’re tracking habits, recurring tasks, or calendared events — use Todoist or similar
Key Concepts
Outcomes and Actions
Bon has exactly two item types. Outcomes are desired results — the “what does done look like?” question. Actions are concrete next steps — the things you actually do. Actions can belong to an outcome, or stand alone. This is GTD’s natural planning model: envision the result, then identify the next physical action.
The Brief
Every item requires a brief with three fields: why (what makes this worth doing), what (the deliverable or change), and done (how you’ll know it’s finished). This isn’t bureaucracy — it’s the minimum context a future session needs to start working without asking “wait, what was this about?” The brief is what makes bon items portable across sessions.
Tactical Steps — The Draw-Down Pattern
When you start working an action (bon work), bon generates tactical steps — a numbered checklist of how to get it done. You work through them one at a time with bon step, which marks the current step complete and shows the next. Only one action may have active tactical steps at a time — this enforces serial execution, the equivalent of working one ticket before grabbing the next. The final bon step auto-completes the action.
This is the draw-down pattern: read the item, activate steps, work with pauses. It prevents the drift that happens when an AI session wanders away from what was actually asked for.
The Draw-Up Pattern
The inverse of draw-down. When you encounter work that shouldn’t be done now, you draw up — file it as a new bon item with a complete brief. The brief quality matters because the session that eventually works this item will have no memory of the conversation where it was filed. Draw-up is how knowledge work accumulates without losing context.
Readiness, Not Urgency
Bon’s list command surfaces items by readiness — what’s actionable right now — rather than by deadline or priority score. bon wait parks an item (and clears its tactical steps). bon unwait brings it back. This is GTD’s core insight applied to AI collaboration: direct attention to what you can act on, not what’s screaming loudest.
How It Relates to Other Tools
Bon tracks what needs doing. The rest of the brigade handles the how:
- Trousse — Trousse’s hooks inject bon’s tactical state into every prompt via
bon-tactical.sh. This means every Claude session automatically knows what step it’s on without anyone having to explain. The/openskill loads bon context at session start. - Aboyeur — The orchestrator reads
bon_hashto detect whether sessions are making progress or stuck. Bon gives aboyeur the ground truth about work state. - Garde-manger — Indexes bon items as a searchable source type, so you can find past work (“did we already do something like this?”) across sessions.
- Mise, Passe, Consomme — These are the tools you reach for while working a bon action. Bon says what; they do the prep, the automation, the analysis.
Design Principle
The human stays in the kitchen. GTD not Agile, readiness not urgency.
Bon embodies this principle most directly. The vocabulary is deliberate — “outcomes” not “epics”, “actions” not “stories”, “waiting” not “blocked”. The human decides what matters and when to engage. The AI works the station.
📦 Install & usage: github.com/spm1001/bon