Grill With Docs

Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.

Ask the questions one at a time, waiting for feedback on each question before continuing.

If a question can be answered by exploring the codebase, explore the codebase instead.

When the user is choosing tools, guardrails, skills, or teaching material, explicitly prompt for long-form rationale before compressing the answer. Ask why this choice, why now, what trade-off is accepted, what a learner/operator should understand, and what future agents should remember. Capture the result in the appropriate durable surface: CONTEXT.md for terms, ADRs for hard-to-reverse decisions, project Brain/docs for source-backed rationale, or skills/TA notes for reusable guidance.

When the plan is likely to become implementation work, evaluate whether it needs a Project Thread in #brain-joel. Recommend creating one when the work crosses any of these thresholds:

• it will produce an ADR, PRD, or multi-step implementation plan
• it will require milestone updates, operator approvals, or canary evidence
• it touches public channel behavior, gateway routing, deploys, secrets, customer-facing systems, or durable runtime state
• it is expected to outlive the current chat turn or involve follow-up verification

Ask for operator approval before creating or using a Project Thread. Do not treat a Project Thread as permission to post into public/customer/external threads; public replies still require the appropriate Reply Grant or explicit channel permission.

Long-form rationale capture

Do not accept a bare preference like "use this library" when the session is meant to teach or bank reusable agent behavior. Grill for:

• the practical reason for the choice
• the alternatives being rejected
• the trade-offs and risks
• the kind of loop, gate, or workflow the choice supports
• whether the rationale is for humans, agents, or both
• where the rationale should live so future agents can use it

Keep the user's full explanation long enough to preserve nuance, then write a compact version only after the durable source exists.

For post-hoc questions like "why did we add all this?", do not lead with a file inventory or "because you asked." Lead with the captured rationale, then map that rationale to files, commands, or guardrails as evidence.

Domain awareness

During codebase exploration, also look for existing documentation:

File structure

Most repos have a single context:

/
├── CONTEXT.md
├── docs/
│   └── adr/
│       ├── 0001-event-sourced-orders.md
│       └── 0002-postgres-for-write-model.md
└── src/

If a CONTEXT-MAP.md exists at the root, the repo has multiple contexts. The map points to where each one lives:

/
├── CONTEXT-MAP.md
├── docs/
│   └── adr/                          ← system-wide decisions
├── src/
│   ├── ordering/
│   │   ├── CONTEXT.md
│   │   └── docs/adr/                 ← context-specific decisions
│   └── billing/
│       ├── CONTEXT.md
│       └── docs/adr/

Use the project’s existing canonical glossary and ADR location. In a Brain project, write .svx under its established structure. The CONTEXT.md layout below is a fallback for projects that use it, not a reason to create a competing truth store. Create only artifacts needed for resolved decisions.

During the session

Challenge against the glossary

When the user uses a term that conflicts with the existing language in CONTEXT.md, call it out immediately. "Your glossary defines 'cancellation' as X, but you seem to mean Y — which is it?"

Sharpen fuzzy language

When the user uses vague or overloaded terms, propose a precise canonical term. "You're saying 'account' — do you mean the Customer or the User? Those are different things."

Discuss concrete scenarios

When domain relationships are being discussed, stress-test them with specific scenarios. Invent scenarios that probe edge cases and force the user to be precise about the boundaries between concepts.

Cross-reference with code

When the user states how something works, check whether the code agrees. If you find a contradiction, surface it: "Your code cancels entire Orders, but you just said partial cancellation is possible — which is right?"

Update CONTEXT.md inline

When a term is resolved, update CONTEXT.md right there. Don't batch these up — capture them as they happen. Use the format in CONTEXT-FORMAT.md.

Don't couple CONTEXT.md to implementation details. Only include terms that are meaningful to domain experts.

Use Project Threads for operator-facing workrooms

A Project Thread is the operator-facing Slack workroom for a bounded objective, usually in #brain-joel. Use it for milestone updates, blockers, audit receipts, canary evidence, and handoff links.

During a grill, if the emerging plan crosses the Project Thread threshold, ask one question before implementation:

Recommended: create/use a #brain-joel Project Thread for this objective so milestones and evidence stay out of public/incident threads. Approve?

If approved, create or reuse a #brain-joel Project Thread. When creating one, the root message must mention Joel directly (<@U030BJ3CK>) and state the bounded objective. Then carry the Project Thread URL/ID into the PRD (project_thread_url) or working notes. Post milestones, blockers, approval requests, and canary evidence there. If declined, continue in the current conversation and keep updates concise.

Offer ADRs sparingly

Only offer to create an ADR when all three are true:

1. Hard to reverse — the cost of changing your mind later is meaningful
2. Surprising without context — a future reader will wonder "why did they do it this way?"
3. The result of a real trade-off — there were genuine alternatives and you picked one for specific reasons

If any of the three is missing, skip the ADR. Use the format in ADR-FORMAT.md.