design-readiness-check（设计准备情况检查）

Design Readiness Check

Overview

This skill is the quality gate between design work and implementation planning.

Its job is to answer one question clearly: is the current design complete enough to move into writing-plans?

It is not a replacement for design-decision-audit. That skill audits standalone design or plan documents. This skill checks the readiness of an in-progress design workflow before planning.

When to Use

Use this skill when:

• the design looks mostly complete
• the next possible step is writing-plans
• the user asks whether the design is ready to implement
• the remaining concern is completeness, not discovery

Do not use this skill when:

• the design tree is still missing
• major branches remain vague
• core decisions are still unresolved
• the user is asking for a general audit of an external design document
• design_target_type is missing from the design state

Workflow

Phase A: Context Preparation

Goal: Load design state and assemble context for parallel checks.

1. Load the current design_state from the conversation context.
2. Read the design tree content and all related context (open branches, decision nodes, risks, validation state).
3. If design_state does not exist or the design tree is empty, return NOT READY immediately with a handoff to design-structure.
4. If design_target_type is missing, return NOT READY immediately with a blocking issue for missing required state.
5. Assemble a context package containing: design tree text, open branches, decision nodes, existing risks, validation entries, and design_target_type.
6. Perform a lightweight structural integrity check using the shared rules in ../design-tree-core/REFERENCE.md:

- mixed responsibilities inside one tree - parent/child ownership confusion - duplicated parent/child logic on the same branch - branches that likely should have been derived but were kept inline

Phase B: Parallel Readiness Checks

Goal: Run four independent checks in parallel using specialized subagents.

1. Launch four parallel Sonnet subagents, each using one of the checker agent templates:

- branch-checker: reads skills/design-readiness-check/agents/branch-checker.md. Fill {{DESIGN_TREE}} and {{CONTEXT}}. - assumption-checker: reads skills/design-readiness-check/agents/assumption-checker.md. Fill {{DESIGN_TREE}} and {{CONTEXT}}. - failure-checker: reads skills/design-readiness-check/agents/failure-checker.md. Fill {{DESIGN_TREE}} and {{CONTEXT}}. - risk-checker: reads skills/design-readiness-check/agents/risk-checker.md. Fill {{DESIGN_TREE}} and {{CONTEXT}}.

1. Collect results from all four subagents.

Fallback: If any subagent fails or times out, the main agent performs that check inline using the same rubric from the agent template.

Phase C: Verdict Synthesis

Goal: Combine check results into a clear readiness judgment.

1. Map each subagent's status to a pass/fail entry in the readiness checklist:

- required-state check → "Design target type present" - branch-checker → "Design tree present" and "Key branches refined" - assumption-checker → "Decisions resolved" (if assumptions are unresolved) - failure-checker → "Failure paths documented" and "Validation strategy defined" - risk-checker → "Blocking risks mitigated" - structural integrity check → "Structural integrity preserved"

1. Build the ✓/✗ checklist diagram following Diagram Conventions.
2. Determine verdict:

• READY: required state is present, all four checks return pass, and no structural integrity issue is blocking
• NOT READY: required state is missing, any check returns fail, or structural integrity issues are blocking

1. If NOT READY, determine the handoff target based on which check failed:

- required-state fails → hand off to design-structure - branch-checker fails → hand off to design-structure (branches missing) or design-refinement (branches weak) - assumption-checker fails → hand off to design-refinement (assumptions need expansion) - failure-checker fails → hand off to design-refinement (failure paths need adding) - risk-checker fails → hand off to decision-evaluation (unresolved risk decision) or design-refinement (risk documentation weak) - structural integrity fails → hand off to design-orchestrator (re-route ownership), design-structure (derive a child tree), or design-refinement (shrink duplicated inline logic)

1. Update design_state with:

- status.ready_for_planning: true or false - status.blocking_issues: list from failed checks - Updated open_branches, risks, validation if new information emerged - If a persisted design artifact exists, keep its document status aligned with the verdict: draft for NOT READY, ready-for-planning for READY

1. Return the explicit verdict with checklist diagram. Never give a "probably ready" answer.

Readiness Standard

A design is ready for planning only when:

• design_target_type is present
• the main design tree is present
• key branches are refined enough to guide implementation for the current target type
• major decision nodes are resolved or explicitly deferred with acceptable rationale
• failure paths and validation strategy are not missing
• blocking risks are either mitigated or clearly acknowledged
• tree responsibilities are not mixed in a way that breaks routing clarity

Expected Outputs

Produce or update a design_state that includes:

• design_target_type
• open_branches
• risks
• validation
• status.ready_for_planning
• status.blocking_issues

Always return an explicit result:

• ready for planning
• not ready for planning

If a persisted design file is part of the workflow, its document status should also be explicit:

• draft while the design is not ready
• ready-for-planning after the readiness gate passes

Diagram Conventions

Present the readiness verdict as a status checklist inside a code block (no language tag):

Readiness Check
├── Design target type present   ✓
├── Design tree present          ✓
├── Key branches refined         ✓
├── Decisions resolved           ✗ (storage choice pending)
├── Failure paths documented     ✓
├── Validation strategy defined  ✓
├── Blocking risks mitigated     ✗ (migration cutover risk)
└── Structural integrity preserved ✓

Verdict: NOT READY — 2 blocking issues

Rules:

• Use ✓ (pass) and ✗ (fail) markers with inline reason for failures
• Max width: 78 characters
• Always include this checklist in the readiness verdict

Entry and Exit Criteria

Enter when:

• the design is near completion
• the next meaningful step may be implementation planning

Exit when:

• you have issued a clear readiness decision
• you have identified the next route if the design is not ready

Handoff Rules

• Hand off to writing-plans only when the design is clearly ready.
• Hand off to design-structure when foundational branches are still missing.
• Hand off to design-refinement when important branches exist but are still weak.
• Hand off to decision-evaluation when the real blocker is an unresolved decision node.
• Hand off to design-orchestrator when the real blocker is tree ownership or routing drift rather than branch weakness.
• Treat missing design_target_type as an immediate blocker, not as a soft warning.
• Never give a "probably ready" answer. The result must be explicit.