Skip to content

Skills Reference

MeowKit ships skills using the mk: namespace. Each skill loads on demand — agents activate only skills relevant to their phase. Detailed procedures live in references/ within each skill directory and load only when needed.

Phase 0 — Orient

Skills for task classification, context loading, and agent routing.

SkillWhat it does
mk:agent-detectorAuto-detect agent, complexity, and model tier per message
mk:lazy-agent-loaderOn-demand agent loading for token efficiency
mk:scoutParallel codebase exploration via Explore subagents
mk:scale-routingDomain-based complexity classification (CSV-driven)
mk:project-contextGenerate/update docs/project-context.md from codebase

Phase 1 — Plan

Skills for planning, research, and ideation.

SkillWhat it does
mk:plan-creatorAuto-select plan template by task scope. 9-step workflow (00-08)
mk:validate-plan8-dimension plan quality check before Phase 2
mk:plan-ceo-reviewCEO-mode plan review (product lens + engineering lens)
mk:office-hoursYC-style brainstorming (startup + builder modes)
mk:brainstormingStructured ideation with scoring and plan-creator handoff
mk:prompt-enhancerRefine a draft user prompt — default emits only the rewrite. --analyze adds decomposition, detected issues, and suggestions; --score adds a deterministic 1–10 quality rating; --deep adds an opt-in codebase scout
mk:decision-frameworkTriage, escalation, and case management for architecture decisions
mk:api-designREST/GraphQL API design patterns

Phase 2 — Test

Skills for test writing, quality gates, and security scanning.

SkillWhat it does
mk:testingRed-green-refactor TDD workflow, validation, visual QA
mk:nyquistTest-to-requirement coverage mapping
mk:lint-and-validateAuto-run linters and type checks
mk:csoChief Security Officer — infrastructure-first audit
mk:vulnerability-scannerOWASP 2025 code-level scanning
mk:skill-template-secureSecure template for skills handling untrusted input

Phase 3 — Build

Skills for implementation, debugging, and domain-specific patterns.

SkillWhat it does
mk:cookEnd-to-end feature pipeline: plan → test → build → review → ship
mk:harnessAutonomous green-field build pipeline with generator/evaluator loop
mk:developmentCode patterns, TDD discipline, skill loading reference
mk:fixStructured bug investigation with auto-complexity detection
mk:investigateSystematic 5-phase root cause debugging
mk:sequential-thinkingHypothesis-driven reasoning for root cause analysis
mk:problem-solvingStrategic unsticking — 7 techniques for "stuck on approach"
mk:simplifyMandatory over-engineering removal pass after build
mk:verifyUnified build→lint→test→type-check→coverage verification
mk:build-fixBuild error triage with language detection and fix references
mk:sprint-contractNegotiate testable acceptance criteria before generator touches code
mk:clean-codePragmatic coding standards (SRP, DRY, KISS)
mk:bootstrapApplication scaffold for any stack with progressive generation

Frontend

SkillWhat it does
mk:typescriptStrict TS: null handling, type guards, utility types
mk:vueVue 3 Composition API, Pinia, reactivity
mk:angularAngular v20+ patterns — components, signals, DI, forms, routing, SSR
mk:react-patternsReact/Next.js performance — 45+ rules from Vercel Engineering
mk:frontend-designUI/UX with anti-AI-slop enforcement
mk:ui-design-systemDesign intelligence: styles, WCAG 2.1 AA, palettes, checklists

Backend & Data

SkillWhat it does
mk:databaseSchema design, safe migrations, query optimization (PostgreSQL primary)
mk:api-designREST/GraphQL patterns: naming, status codes, pagination, versioning

Phase 4 — Review

Skills for code review, security audit, and elicitation.

SkillWhat it does
mk:reviewMulti-pass adversarial code review with parallel reviewers
mk:elicitStructured second-pass reasoning via 8 methods (pre-mortem, red team, Socratic)
mk:evaluateBehavioral active-verification grader with rubric presets
mk:rubricLoad, validate, and compose rubric presets for evaluator grading

Phase 5 — Ship

Skills for deployment and release management.

SkillWhat it does
mk:shipShip pipeline: test → review → commit → PR
mk:worktreeGit worktree isolation for parallel agent sessions

Phase 6 — Reflect

Skills for documentation, retrospectives, and memory capture.

SkillWhat it does
mk:memorySession capture, pattern extraction, cost tracking
mk:retroSprint retrospective with commit analysis and trend tracking
mk:document-releasePost-ship doc sync — cross-reference diff against docs
mk:docs-initGenerate initial docs from codebase analysis

Cross-Phase Skills

Skills that can activate in any phase.

SkillWhat it does
mk:docs-finderLibrary/project docs via Context7 + Context Hub
mk:helpNavigation assistant — scans project state, recommends next pipeline step
mk:carefulWarn before rm -rf, DROP TABLE, force-push
mk:freezeRestrict edits to one directory during debugging
mk:multimodalImage/video/audio/PDF analysis via Gemini API
mk:agent-browserChrome/CDP with session persistence and auth
mk:web-to-markdownConvert web pages to markdown for context injection
mk:llmsGenerate llms.txt per llmstxt.org spec
mk:session-continuationSave/resume workflow state across sessions
mk:henshinWrap existing code as agent-consumable surfaces (CLI + MCP + skill)
mk:intakeTicket analysis, completeness scoring, product area detection
mk:task-queueQueue management for multi-task sessions

QA & Browser

SkillWhat it does
mk:qaSystematic QA with bug fixing and health scores
mk:qa-manualSpec-driven manual QA + Playwright E2E code generation
mk:playwright-cliPlaywright MCP with code generation

External Integrations

SkillWhat it does
mk:jiraRouting hub for the 16-leaf Jira family backed by the jira-as CLI
mk:confluenceRouting hub for the 5-leaf Confluence Cloud family (page CRUD, CQL search, spec analysis with macro-aware ADF fetch, bulk ops, collaboration) backed by the confluence-as CLI
mk:figmaFigma design analysis + implementation via Figma MCP. 3 modes: analyze, implement, tokens

Infrastructure

SkillWhat it does
mk:skill-creatorScaffold, validate, and register new skills
mk:project-organizationFile naming and directory structure standards
mk:team-configTeam-level MeowKit configuration

Special-Purpose

SkillWhat it does
mk:partyMulti-agent deliberation for architecture decisions (2-4 agents, forced synthesis)
mk:workflow-orchestratorAuto-invoked 7-phase orchestrator for complex-feature intent
mk:benchmarkCanary suite for dead-weight audit measurement
mk:trace-analyzeScatter-gather analysis of harness trace logs
mk:planning-engineCodebase-aware tech breakdown and sprint planning analysis
mk:chomCode chomping — extract relevant code sections for context
mk:packPackage code for portable context injection

Plan-first gate

Most pipeline skills enforce a plan-first gate. Skills that skip planning either produce planning input or are data-driven.

SkillGate behaviorSkip condition
mk:cookCreate plan if missingPlan path arg, --fast mode
mk:fixPlan if > 2 files--quick mode
mk:shipRequire approved planHotfix with human approval
mk:csoScope audit via plan--daily mode
mk:reviewRead plan for contextPR diff reviews

See also

Released under the MIT License.