agentsclimarketplace

Dev kit

Skill SoftwareOneHN/dev-kit

πŸ› οΈ Dev Kit β€” lifecycle-driven Agent Skill for structured development workflows. Brainstorm, plan, implement, debug, fix, test & adversarial code review. Java/Spring Boot + TypeScript/React/Vue rules included. Compatible with Claude Code, Cursor, Kiro, and 40+ agentic clients.

Install
npx -y skills add SoftwareOneHN/dev-kit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Full-stack Development Kit β€” structured workflows for brainstorming, planning, implementation, debugging, testing, and code review. Use when the user mentions: plan, implement, code, cook, fix, debug, test, review, scout, brainstorm, feature, bug, refactor, architecture, pipeline, CI, performance, root cause, or asks to build/ship/deploy features.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

7.6 KB, as published. Nobody here has run it

Dev Kit

Structured development workflows β€” from ideation to production, with adversarial review and root cause analysis at every stage.

Dev Lifecycle

brainstorm β†’ plan β†’ cook (implement) β†’ test β†’ code-review β†’ ship
                         ↕
                    debug ↔ fix

Key Capabilities

  • Structured brainstorming with trade-off analysis and exact requirements extraction
  • Implementation planning with research phases and solution design
  • Feature implementation pipeline (scout β†’ plan β†’ code β†’ test β†’ review β†’ finalize)
  • Systematic debugging with 10 technique references (root cause, logs, CI, perf, frontend)
  • Bug fixing with complexity routing, prevention gates, and no-side-effect verification
  • Test execution (unit/integration/e2e/UI) with coverage and QA reports
  • Adversarial 3-stage code review (spec compliance β†’ quality β†’ red-team)
  • Fast codebase scouting with parallel agents

Routing β€” Intent to Skill

When the user says something, route to the correct skill:

User IntentSkillReference
Brainstorm ideas, evaluate approachesbrainstormreferences/brainstorm/SKILL.md
Plan implementation, design architectureplanreferences/plan/SKILL.md
Implement feature, execute plancookreferences/cook/SKILL.md
Fix bug, resolve errorfixreferences/fix/SKILL.md
Debug issue, investigate root causedebugreferences/debug/SKILL.md
Run tests, check coveragetestreferences/test/SKILL.md
Review code, audit PRcode-reviewreferences/code-review/SKILL.md
Find files, explore codebasescoutreferences/scout/SKILL.md

Priority when ambiguous:

  • Prefer fix if user mentions a concrete error or failing test
  • Prefer debug if user mentions investigation, root cause, or "why"
  • Prefer cook if user mentions implementing or building something new
  • Prefer brainstorm if user says "should we", "what if", or "how about"
  • Prefer plan if user mentions architecture, roadmap, or phases

Defaults

  • Language: Vietnamese for explanations, English for code
  • Principles: YAGNI, KISS, DRY β€” always
  • Scout first: always scan codebase before asking questions or forming hypotheses
  • No guessing: evidence-based decisions only
  • Verify before claiming: fresh test output required before any "done" claim

Behavior Rules

Scout-First (Mandatory)

Every skill MUST scan the codebase before:

  • Asking clarifying questions
  • Forming hypotheses
  • Proposing approaches
  • Writing code

No Side Effects

Implementation and fixes are NOT done until verified side-effect-free:

  1. All tests pass (modified + transitively-affected modules)
  2. No business logic regression in blast radius
  3. No new lint/type/build errors
  4. Public contracts unchanged unless intentional

Adversarial Review (Always-On)

Every code change gets 3-stage review:

  1. Spec compliance β€” does it match what was requested?
  2. Code quality β€” standards, security, performance
  3. Red-team β€” actively tries to break the code

Verification Iron Law

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE. Run command β†’ Read output β†’ Confirm pass β†’ Then claim result.

Skill Details

brainstorm

Structured ideation with scout-first, exact requirements gates, and trade-off analysis. Ends with design doc + handoff to /plan.

plan

Implementation planning with research phase, codebase understanding, solution design, and phase-by-phase output. Creates actionable plans for /cook.

cook

End-to-end feature implementation. Modes: --interactive (default), --fast, --auto, --parallel, --no-test, --tdd. Full pipeline: scout β†’ research β†’ plan β†’ implement β†’ test β†’ review β†’ finalize.

fix

Bug fixing with intelligent routing. Modes: --auto (default), --review, --quick, --parallel. Pipeline: scout β†’ diagnose β†’ assess complexity β†’ fix β†’ verify+prevent β†’ finalize.

debug

Systematic debugging framework. Techniques: systematic debugging, root cause tracing, defense-in-depth, verification, investigation methodology, log/CI analysis, performance diagnostics, frontend verification.

test

Comprehensive testing. Supports: unit/integration/e2e (Jest/Vitest/pytest/Go/Rust/Flutter), UI testing via browser automation, coverage analysis, structured QA reports.

code-review

Adversarial review with input modes: #PR, commit hash, --pending, codebase, codebase parallel. Three-stage protocol + task-managed review pipelines for multi-file features.

scout

Fast codebase scouting using parallel agents. Supports internal (Explore subagents) and external (Gemini/OpenCode) modes. Divides codebase into segments for maximum coverage.

Language-Specific Rules

Load the appropriate reference when working with a specific language/framework:

LanguageReferenceWhen to Load
Java (core)references/languages/java/core-rules.mdAny Java code β€” bans, musts, style, error handling
Java (Spring Boot)references/languages/java/spring-boot.mdSpring controllers, services, repos, config, exceptions
Java (Clean Arch)references/languages/java/clean-architecture.mdLayer structure, dependency rules, anti-patterns
Java (Testing)references/languages/java/testing.mdJUnit 5, Testcontainers, Mockito, JPA testing
TypeScript (core)references/languages/typescript/core-rules.mdAny TS/frontend code β€” bans, type safety, a11y, security
TypeScript (React/Next.js)references/languages/typescript/react-nextjs.mdReact components, Zustand, TanStack Query, App Router
TypeScript (Vue/Nuxt)references/languages/typescript/vue-nuxt.mdVue 3 Composition API, Pinia, Nuxt patterns
TypeScript (API)references/languages/typescript/api-patterns.mdFetch wrapper, BFF, response validation, optimistic updates
TypeScript (Forms)references/languages/typescript/forms.mdReact Hook Form, VeeValidate, Zod schemas, Server Actions

Shared References

  • references/_shared/workflow-artifacts.md β€” Review artifact schema and validator contract

Tech Stack Support

Language/framework agnostic. Works with any:

  • Languages: TypeScript, JavaScript, Python, Go, Rust, Java, Flutter/Dart
  • Frameworks: React, Next.js, Express, FastAPI, Django, Spring, etc.
  • Test runners: Jest, Vitest, Mocha, pytest, go test, cargo test, flutter test
  • CI/CD: GitHub Actions, GitLab CI, any pipeline
  • Tools: gh CLI, git, npm/pnpm/yarn, pip, cargo, go

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.