grill-with-docs
Grilling session that challenges your plan against the existing domain model, sharpens terminology, and updates documentation (CONTEXT.md, ADRs) inline as decisions crystallise. Use when user wants to stress-test a plan against their project's language and documented decisions.
Security Assessment
About grill-with-docs
A Simplified-Chinese grilling skill that stress-tests a plan by interrogating it against a project's existing domain model and documented decisions, sharpening terminology and updating documentation inline as decisions crystallize. It is used when someone wants their plan challenged against their project's own language and recorded decisions. The interaction style is deliberate: it asks one question at a time and waits for feedback before continuing, works down each branch of the design tree resolving dependencies between decisions, always offers its own recommended answer, and, when a question can be answered by exploring the codebase, investigates the code instead of asking.
It builds domain awareness from existing documentation. In most repos that means a single CONTEXT.md glossary plus architecture decision records under docs/adr/; when a CONTEXT-MAP.md exists at the root, the repo has multiple bounded contexts, each with its own CONTEXT.md and ADR folder, and the map points to them. Files are created lazily, a CONTEXT.md only when the first term is resolved and an ADR folder only when the first ADR is needed.
During a session it applies several tactics. It challenges terminology that conflicts with the existing glossary, for example when the user's meaning of cancellation differs from the documented definition; it sharpens fuzzy or overloaded terms toward a precise canonical one, such as distinguishing account as Customer versus User; and it stress-tests domain relationships with concrete, invented scenarios that probe edge cases and boundaries between concepts. It cross-references explanations against the code and points out contradictions, such as code that cancels whole orders when the user claims partial cancellation is possible. Resolved terms are captured in CONTEXT.md immediately using a defined format, and the glossary is kept purely as a glossary with no implementation detail, specs, or scratch notes. ADRs are proposed sparingly, only when a decision is hard to reverse, surprising without context, and the result of a real trade-off; if any of those three is missing, it skips the ADR. Dedicated format files (CONTEXT-FORMAT.md, ADR-FORMAT.md) define the output.
FAQ
How does the questioning work?
It asks one question at a time and waits for your feedback before moving on, walking each branch of the design tree, and it always provides its own recommended answer for each question.
Where does it get the project's domain knowledge?
From existing docs, typically a CONTEXT.md glossary and ADRs under docs/adr/; if a CONTEXT-MAP.md exists at the root, the repo has multiple contexts, each with its own CONTEXT.md and ADR folder.
When does it create documentation files?
Lazily. It creates CONTEXT.md only when the first term is resolved and a docs/adr/ folder only when the first ADR is needed, capturing content as it happens rather than at the end.
What belongs in CONTEXT.md?
Only the glossary. It must contain no implementation details and should not be treated as a spec, scratch pad, or decision repository.
When will it propose an ADR?
Only when all three are true: the decision is hard to reverse, surprising without context, and the result of a real trade-off; if any one is missing, it skips the ADR.
Install grill-with-docs
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
vinvcn/mattpocock-skills-zh-cn