solution-review

name	solution-review
version	1.1.0
user-invocable	true
description	Authoritative framework for reviewing any proposed solution (code, config, process, tooling, or architecture decision) before implementation. Covers four core dimensions (effectiveness, side-effects/risks, feasibility, spec compliance) plus five strategic dimensions that catch decision-level failures the core four miss: reversibility calibration (Bezos one-way/two-way doors), failure-mode analysis (FMEA), operability (SRE Production Readiness Review), cost-vs-value (CBAM/WSJF), and team cognitive fit (Team Topologies). Use this skill whenever you need to review, validate, stress-test, or approve a proposed solution, plan, design doc, or approach — before any code is written or change is applied. Triggers — 「审查方案」「方案审查」「评估方案」「方案评估」「review proposal」「方案可行吗」「这个方案靠谱吗」「方案有没有风险」「方案评审」「设计评审」「决策评审」 / solution review, proposal review, design review, decision review, stress-test a plan, validate an approach.

Solution Review

Role: A pre-implementation review framework for any proposed solution — code or non-code. It exists to catch decision-level failures (wrong problem, uncalibrated risk, ignored operability, poor cost-value tradeoff) before effort is spent building. It is a methodology enhancement: PDCA hosts declare this skill in frontmatter dependencies and invoke it after a prerequisite check (the host aborts at startup if this skill is missing). It complements code-design-review (which handles code-level design quality when the solution involves code).

Detailed framework entries (author, year, method, source) and the full blocking/non-blocking criteria live in reference.md.

Why this skill exists

Most solution reviews stop at four questions: Does it solve the problem? Any side effects? Is it feasible? Does it follow conventions? These are necessary but not sufficient. They catch implementation-level problems but miss decision-level failures:

• A solution that solves the right problem but is irreversible when it did not need to be (a one-way door treated as a two-way door).
• A solution whose failure modes were never enumerated, so the first production incident is a surprise.
• A solution that works but is un-operable — no monitoring, no rollback path, on-call burden nobody accounted for.
• A solution whose cost exceeds its value but nobody ran the numbers.
• A solution that is technically correct but exceeds the team's cognitive capacity to maintain.

This skill adds five strategic dimensions that surface these failures, calibrated to the solution's reversibility. The meta-lesson:

A solution can pass all four core dimensions and still be the wrong decision. Review depth must scale with irreversibility, not with confidence.

When to use

Strong signals (any one):

• You are about to approve, reject, or iterate on a proposed solution, plan, or design — before implementation begins.
• A solution involves an architecture decision, a migration, a dependency change, or an irreversible action (schema change, data migration, public API contract).
• Someone asks "is this approach sound?" or "what could go wrong?" or "should we do this?"
• You need to produce a review report with structured pass/fail reasoning, not a gut feeling.

This skill reviews proposed solutions (pre-implementation). It is distinct from post-implementation code review skills (which review completed diffs against standards). Both matter; they run at different lifecycle stages.

Relationship to other skills

Skill	Responsibility	Relationship to this skill
code-design-review	Code-level design quality (coupling, cohesion, complexity, security)	When the solution involves code, use both: this skill checks the decision; code-design-review checks the code craft. This skill is the superset.
PDCA hosts	Host workflows	Referenced by PDCA hosts via frontmatter dependencies (missing → host aborts). Deep-dive of their "review solution" phase. Hosts are SoT for that edge.

The review framework

Nine dimensions, in two tiers. The four core dimensions are mandatory for every solution. The five strategic dimensions are mandatory for solutions above a reversibility threshold (see dimension 5) and recommended for all.

Core dimensions (every solution, every time)

1. Effectiveness — Does the solution completely cover the root cause / requirement? Is the core logic sound? Are there gaps where the problem persists after the solution is applied?
2. Side-effects & risks — What breaks elsewhere (functional side-effects)? What non-functional regressions are possible (performance, security, maintainability)? Are identified risks paired with mitigation measures?
3. Feasibility — Are the change scope, dependencies, and affected files/modules concrete and actionable? Can this actually be built with current constraints (time, budget, knowledge, infrastructure)?
4. Spec & standards compliance — Does the solution match the originating requirement / spec / PRD? Does it follow existing project patterns and documented conventions?

Strategic dimensions (calibrate depth by reversibility — see dim 5)

5. Reversibility calibration (Bezos, Type 1 vs Type 2 Decisions, 2016 Amazon Shareholder Letter) — Is this a one-way door (irreversible: data migration, public API, schema change) or a two-way door (easily rolled back)? One-way doors demand maximum review rigor; two-way doors favor speed. Misclassifying a one-way door as two-way is the most expensive review error.
6. Failure-mode analysis (FMEA — Failure Mode and Effects Analysis, ASQ/NASA) — Enumerate how the solution can fail: for each component, what failure modes exist? What is the severity × likelihood × detectability? High-Risk-Priority-Number failures need mitigation before approval.
7. Operability (Google SRE, Production Readiness Review, 2016) — If this ships, can it be operated? Are there SLIs/SLOs? Is there monitoring and alerting? Is there a rollback path? What is the on-call burden? A solution that works but cannot be operated is a future incident.
8. Cost vs value (CBAM — SEI/CMU, Cost Benefit Analysis Method, 2003; and WSJF — Reinertsen, Principles of Product Development Flow, 2009) — Is the effort worth the outcome? What is the cost of delay? Is there a cheaper alternative that achieves 80% of the value? Include long-term maintenance / change-amplification cost. Low implementation effort (including AI-assisted coding) does not by itself justify a structurally weaker design when a better feasible alternative exists.
9. Team cognitive fit (Skelton & Pais, Team Topologies, 2019) — Does the solution fit the team's current cognitive capacity? Does it require knowledge the team does not have and cannot quickly acquire? A technically superior solution the team cannot maintain is a liability, not an asset.

Detailed method, application steps, and source citations for each dimension are in reference.md.

How to apply

1. Classify reversibility first. Before reviewing anything else, determine: is this a one-way door or a two-way door? This sets the review depth for all subsequent dimensions. (See reference.md → Reversibility.)
2. Run the four core dimensions. These are non-negotiable for any solution. Produce a pass/fail for each with concrete reasoning.
3. Run the five strategic dimensions at the depth the reversibility classification demands. For two-way doors, a quick pass is sufficient; for one-way doors, each strategic dimension deserves a dedicated assessment.
4. Produce a structured review report (see Output below) with: per-dimension verdict, identified issues ranked by severity, blocking vs non-blocking classification, and an overall conclusion.
5. Gate on blocking issues. Any blocking issue means the solution does not pass. Non-blocking issues are noted as recommendations but do not block.

Blocking vs non-blocking (summary)

Blocking (any one → solution does not pass):

• The solution does not completely cover the root cause / requirement (effectiveness gap).
• An identified medium/high risk has no mitigation measure.
• The change scope or dependencies are unclear — no executable plan can be derived.
• The solution conflicts with existing design patterns or coding conventions.
• A likely side-effect could introduce new bugs or break existing functionality, unhandled.
• (One-way doors only) A failure mode with high severity × high likelihood has no mitigation.
• (One-way doors only) The solution has no rollback path and is irreversible.
• (One-way doors only) The cost clearly exceeds the value and no cheaper alternative was considered.
• A structural / architectural alternative is clearly superior for long-term maintainability, is feasible in scope, and is rejected only because the weaker option is faster to implement — without Prudent-Deliberate debt acceptance. (When the solution involves code, align with code-design-review architecture gates.)

Non-blocking (note as recommendation, do not block):

• Low-risk items that already have mitigation.
• Style preferences that do not affect correctness or structure.
• Performance improvements deferrable to a later iteration (when not structural).
• Style-level elegance without structural / change-amplification impact.
• Deliberate, documented technical debt with a repayment plan (including explicitly accepted weaker architecture).

The full blocking/non-blocking criteria with framework-specific thresholds are in reference.md.

Output

A review report with this structure:

【Solution review report】
- Subject: [solution name / summary]
- Reversibility: [one-way door / two-way door] → review depth: [maximum / standard]
- Core dimensions:
  1. Effectiveness: ✅/❌ [reasoning]
  2. Side effects & risks: ✅/❌ [reasoning]
  3. Feasibility: ✅/❌ [reasoning]
  4. Spec compliance: ✅/❌ [reasoning]
- Strategic dimensions (by reversibility depth):
  5. Reversibility calibration: [classification + rationale]
  6. Failure-mode analysis: ✅/❌/N/A [top failure modes + RPN]
  7. Operability: ✅/❌/N/A [monitoring/rollback/on-call assessment]
  8. Cost vs value: ✅/❌/N/A [cost-value assessment]
  9. Team cognitive fit: ✅/❌/N/A [team fit assessment]
- Issue list: [#] [description] [severity: blocking/non-blocking]
- Verdict: ✅ pass / ❌ fail
- [On ❌] Optimization guidance: [per-issue improvement direction]

Anti-patterns (forbidden moves)

1. Skipping strategic dimensions for non-code solutions — config/process/tooling changes can be one-way doors too (a migration run is irreversible; a process change is hard to undo). Reversibility, not solution type, determines depth.
2. Vague risk assessment — "there might be some risk" without enumerating failure modes, severity, and likelihood. Use the FMEA structure in dimension 6 instead of a gut-feel statement.