원클릭으로
grill-with-docs
Use during discuss-slice and plan-slice to stress-test a spec or plan against domain vocabulary before approval.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Use during discuss-slice and plan-slice to stress-test a spec or plan against domain vocabulary before approval.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when reviewing architecture decisions. C4 model, dependency inversion, hexagonal boundaries.
Use during research-slice to build disposable experiments that answer technical uncertainty.
Use during project:init to scaffold repo-level agent configuration (CLAUDE.md, AGENTS.md, tracker labels).
Use when debugging. 4-phase investigation, root cause, minimal fix.
Use when implementing features/fixes. Iron law: ¬∃ production code without failing test.
Use during discuss-slice to synthesize conversation context into a formal PRD with user stories, implementation decisions, testing decisions, and exclusions.
| name | grill-with-docs |
| description | Use during discuss-slice and plan-slice to stress-test a spec or plan against domain vocabulary before approval. |
| version | 1.0.0 |
| tags | ["process","review"] |
During discuss-slice (stress-test the spec) and plan-slice (stress-test the plan). Validates that proposed designs align with documented vocabulary and boundaries.
Default to light-touch grilling. Escalate to full grill only when:
Load CONTEXT.md, DOMAIN.md, or equivalent. Note defined terms, boundaries, and invariants.
For each claim in the spec or plan:
When nomenclature crystallizes during grilling, add it to the glossary immediately. The glossary must remain devoid of implementation details.
When a decision would be hard to reverse, surprises future maintainers absent context, or emerges from authentic trade-offs, draft a decision record in docs/adr/.
Present gaps, conflicts, and recommended fixes. Do not block approval for minor terminology mismatches — flag them for correction during execution.