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.
Domain awareness
During codebase exploration, also look for existing documentation:
File structure
A project has one glossary, one product spec, and one drift log. Paths are defaults — if the repo already names them differently, its names win. AGENTS.md is cold-boot orientation, not the spec. Discover the spec: docs/INTENT-AND-DESIGN.md if present, else docs/REQUIREMENTS-AND-NOTES.md (greenfield default). Never create the template name just to match this skill. Plain prose is a valid spec. Do not force an architecture-table template onto an existing intent document.
/
├── AGENTS.md ← cold-boot orientation only
├── CONTEXT.md ← glossary only: terms, _Avoid_, _See also_
├── docs/
│ ├── INTENT-AND-DESIGN.md ← existing intent doc, if the repo has one
│ ├── REQUIREMENTS-AND-NOTES.md ← greenfield default only, if no intent doc
│ └── DECISIONS.md ← execution drift log (orchestrator-owned)
└── <source>/
Architectural choices that meet the trade-off test go in the spec. Delivery-time drift goes in the drift log. Do not patch orientation into AGENTS.md to document a product decision.
Create files lazily — only when you have something to write. If CONTEXT.md exists, update it in place and never recreate it; if it does not, create it when the first term resolves, using CONTEXT-FORMAT.md. Patch the spec if present. Start docs/REQUIREMENTS-AND-NOTES.md only on a blank project with no intent document, using SPEC-FORMAT.md.
On a blank slate
If the repo has no code and no CONTEXT.md yet, exactly two of the techniques below have nothing to work with — challenge against the glossary and cross-reference with code. Say so once, then lean on sharpen fuzzy language and discuss concrete scenarios instead, and keep writing the glossary as terms resolve. Do not repeatedly search a codebase that isn't there; cross-referencing resumes as soon as there is code.
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.
CONTEXT.md should be totally devoid of implementation details. Do not treat CONTEXT.md as a spec, a scratch pad, or a repository for implementation decisions. It is a glossary and nothing else.
Record architectural trade-offs in the spec
When a decision meets all three of these, it needs a written home:
- Hard to reverse — the cost of changing your mind later is meaningful
- Surprising without context — a future reader will wonder "why did they do it this way?"
- The result of a real trade-off — there were genuine alternatives and you picked one for specific reasons
Patch the existing spec in its own shape. If the spec already uses an Architecture table, add a row naming the rejected alternative (see SPEC-FORMAT.md). If it is prose, add a short prose decision. Do not convert a prose spec into SPEC-FORMAT.md just to have a table.
If any of the three is missing, write nothing.
Where the grill session ends
A grill session produces a sharpened glossary and a patched spec — it does not produce a PRD. There is no separate requirements document to write: the spec is the artifact, and it is durable rather than per-feature.
When the plan is settled, hand off to the to-issues skill, which reads the spec and glossary directly and breaks the work into issues/.