ayoubben18/ab-method / .agents/skills/grill-with-docs
grill-with-docs
VerifiedGrilling session that challenges your plan against the existing domain model, sharpens terminology, and updates documentation (CONTEXT.md, ADRs) inline as decisions crystallise. Parks the rare question you genuinely can't answer yet as an unresolved question with a placeholder, so the build proceeds instead of stalling. Use when user wants to stress-test a plan against their project's language and documented decisions.
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. Asking multiple questions at once is bewildering.
If a fact can be found by exploring the environment (filesystem, tools, codebase), look it up rather than asking me. The decisions, though, are mine — put each one to me and wait for my answer.
Do not act on the plan until I confirm we have reached a shared understanding.
A branch is closed when it is answered — or, rarely, when it is parked as an unresolved question (see below). Parking is the only sanctioned way to leave a branch open; never leave one silently unresolved.
Before the first question on an area I haven't had explained yet, brief me on it — see Brief before you ask below. When a question has options, explain every option so I can choose without looking anything up — see Presenting choices below.
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/
Create files lazily — only when you have something to write. If no CONTEXT.md exists, create one when the first term is resolved. If no docs/adr/ exists, create it when the first ADR is needed.
During the session
Brief before you ask
Assume the user did not write the code under discussion — an agent did — so every module is new to them until you have explained it in this session. Before the first question on a branch that touches such an area, brief them, then ask. Build the picture in this order:
- The pieces — the 2–4 concepts the decision rests on, each defined in one plain line. Name the domain thing ("a bot joins a workspace with a role"), not the code thing (
grant,#mutate). - How it works today — walk one concrete case from the ticket through the system: who does what, what gets checked, what happens. A small table or step diagram beats prose.
- What's wrong — the symptom the user sees first, then the cause. If it splits into several problems, number them and say which one is the decision and which get fixed whatever they pick.
- What we're deciding — one sentence. Then the question and its options.
Rules:
- Every word in the question must already be explained. A function name, flag, error code or internal term that the briefing did not introduce gets translated or cut. File references (
path:line) are pointers for later, never the explanation. - Size it to the gap. A full briefing for a module not yet explained this session; one line when it was. Never re-explain what is already explained.
- Lead with one line saying what we're about to decide, then the briefing, then the question. This is the one place a grill message doesn't open with the question itself.
See Example 3 in EXAMPLES.md.
Presenting choices
When a question has options to pick from, do this for every option:
- Say what changes — one or two plain sentences on what happens differently for the people and the system: who can now do what, what the user sees, what gets easier or harder, when you'd pick it.
- Then show it in code — a short, simplified snippet that makes the difference visible. Use only names the briefing already introduced; otherwise a tiny made-up example. A few lines, not a full implementation. Skip the snippet when the sentence already makes the difference obvious.
Keep the language simple:
- No vague or buzzword-only terms ("more scalable", "cleaner", "decoupled", "idiomatic"). Say what actually happens instead.
- If a technical term is unavoidable, explain it in the same sentence.
- Make the options differ visibly — side by side, the snippets should make the trade-off obvious.
Then give your recommendation and the one-line reason. See EXAMPLES.md for what this looks like.
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: edit the existing entry in place rather than appending, and run the check-docs gate after writing (.ab-method/rules/docs.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.
Park what the user genuinely cannot answer yet
Occasionally the grill hits a question the user simply cannot close right now — the decision belongs to someone else, depends on data that doesn't exist yet, or is waiting on something outside the codebase. The feature still has to get built. Rather than stalling the grill or letting the agent quietly guess, park the question and build a black box: a named placeholder, generic or empty, that can be swapped when the answer arrives.
This is rare. Expect zero parked questions in most sessions. It is not a valve for "the user hesitated" or "this is getting long".
When to park
All four must hold:
- You asked and recommended. The grill's default is a recommended answer; parking only happens after the user has heard yours and still can't decide.
- The blocker is real and external. They're waiting on a person, a contract, a measurement, a product call — not on thinking about it for another minute. You must be able to name the blocker in one line.
- A wrong guess would be costly or invisible. If any sane default works and nobody would be misled, don't park — take the default, note it in the tracker's Constraints/Notes, and move on. Parking is for when a silently-guessed answer would look deliberate and be wrong.
- The rest of the work survives without it. There is a coherent placeholder that keeps the flow whole. If the mission cannot proceed without inventing data or behaviour, it isn't parkable — say the task is blocked and stop.
Park vs. handoff vs. ADR
- Park (unresolved question) — an open decision inside this task's scope that the build must route around now. Stays with the task.
- Handoff — a tangent that deserves its own task. Different scope, not a black
box. Use the
handoffskill; the current task doesn't wait on it. - ADR — a decision that was made, is hard to reverse, and needs its rationale recorded. The opposite of a parked question. When a parked question is later resolved and meets the ADR bar, write the ADR then.
Never park something that changes the domain language. A term you need to name the
code cannot be deferred — settle it in CONTEXT.md and grill the behaviour behind it.
How to park
- Say it plainly: "Parking that as UQ-2 — we'll ship a fixed 30-day window as a placeholder and swap it when product decides." Get the user's nod on the placeholder, not on the answer.
- Record it in
unresolved-questions.mdnext to the task'sprogress-tracker.md(path from.ab-method/structure/index.yaml), using the format in UNRESOLVED-QUESTIONS-FORMAT.md. If no task folder exists yet, hold the parked questions in the session and write the file the moment it does. - Mark every mission that builds on the black box with
⚠️ UQ-nin the tracker, so whoever runs it knows a placeholder is in play. - Return to the branch you were grilling. Parking one question does not end the grill.
Then keep going as normal — the black box is a placeholder in the plan, not a hole in it.
Offer ADRs sparingly
Only offer to create an ADR when all three are true:
- 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
If any of the three is missing, skip the ADR. Use the format in ADR-FORMAT.md.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |