From klair-legacy
Critically reviews a spec for feasibility, completeness, risks, and shortcomings. Use after spec creation to get an objective, non-sycophantic assessment before implementation. Triggers on "review spec", "spec review", "critique spec", "audit spec".
npx claudepluginhub ai-builder-team/ai-builder-plugin-marketplace --plugin klair-legacyThis skill is limited to using the following tools:
Performs an adversarial, objective review of a technical specification. This is NOT a consistency check (spec-coherence-checker and spec-qc already do that). This skill questions whether the spec is *correct*, *complete*, and *buildable* — catching problems that internal consistency checks miss.
Generates design tokens/docs from CSS/Tailwind/styled-components codebases, audits visual consistency across 10 dimensions, detects AI slop in UI.
Records polished WebM UI demo videos of web apps using Playwright with cursor overlay, natural pacing, and three-phase scripting. Activates for demo, walkthrough, screen recording, or tutorial requests.
Delivers idiomatic Kotlin patterns for null safety, immutability, sealed classes, coroutines, Flows, extensions, DSL builders, and Gradle DSL. Use when writing, reviewing, refactoring, or designing Kotlin code.
Performs an adversarial, objective review of a technical specification. This is NOT a consistency check (spec-coherence-checker and spec-qc already do that). This skill questions whether the spec is correct, complete, and buildable — catching problems that internal consistency checks miss.
/spec or /creating-specs has produced a spec.md/implement or /orchestrate kicks off implementationLocate Spec
│
▼
Read Spec + Research Artifacts
│
▼
Verify Against Codebase (parallel agents)
│
▼
Assess Review Dimensions
│
▼
Produce spec-review.md
│
▼
Present Findings + Verdict
Collect all paths for reference:
{spec-path}/spec.md (required){spec-path}/spec-research.md (optional){spec-path}/checklist.md (optional){feature-path}/FEATURE.md (optional)If spec.md does not exist, stop and inform the user.
The spec makes claims about existing code. Verify them. Launch 3 parallel agents:
Agent 1: File Existence & Structure
subagent_type: "Explore"
description: "Verify spec file references"
prompt: "The following files are referenced in a spec. For each one:
1. Determine which bucket it belongs to:
- Files to Reference: must already exist
- Files to Create/Modify with Action=modify: must already exist
- Files to Create/Modify with Action=create: may not exist yet (expected before implementation)
2. If the file should exist, confirm it exists and verify the interfaces/functions/exports the spec mentions are actually present
3. If Action=create and the file does not exist, treat that as expected (not a defect)
4. If the action is unclear, call that out as a mismatch in the spec
Files referenced:
{list from spec's 'Files to Reference' and 'Files to Create/Modify' sections}
Report format:
- CONFIRMED: {file} — {what was verified, or "planned create; not present yet (expected)"}
- MISSING: {file} — expected existing file does not exist (reference file or modify target)
- MISMATCH: {file} — exists but {what differs from spec's claims}, OR action/create-modify intent is ambiguous"
Agent 2: Interface Contract Validation
subagent_type: "Explore"
description: "Validate interface contracts"
prompt: "The spec defines these interface contracts:
{paste interface contracts from spec}
Search the codebase to verify:
1. Do the types/interfaces referenced actually exist with the fields described?
2. Do the API endpoints return the fields the spec claims?
3. Are there fields the spec assumes exist but don't?
Report CONFIRMED, MISSING, or MISMATCH for each contract element."
Agent 3: Pattern Consistency
subagent_type: "Explore"
description: "Check pattern consistency"
prompt: "The spec proposes this technical approach:
{paste technical design summary}
Search the codebase for similar features. Report:
1. Does this approach match how similar things are built in this codebase?
2. Are there established patterns the spec ignores or contradicts?
3. Are there utilities, hooks, or helpers the spec should use but doesn't mention?"
Wait for all agents. Collect findings.
Using the spec content AND the codebase verification results, evaluate each dimension from guidelines/review-dimensions.md.
For each dimension, produce:
| Severity | Meaning | Action Required |
|---|---|---|
| Critical | Will cause implementation failure or incorrect behavior. Must fix before implementing. | Block implementation |
| Major | Significant gap that will cause rework or missed requirements. Should fix. | Fix recommended |
| Minor | Small issue that won't block implementation but should be addressed. | Fix if convenient |
| Suggestion | Improvement idea, not a defect. Take it or leave it. | Optional |
Write the review report to {spec-path}/spec-review.md using the template at templates/spec-review-template.md.
Do NOT: