brainstorming → technical-documentation → conformance-check → reconciling-changes → planning → project-setup → executing-plans → shipping
Agent skills
Skills
A collection of agent skills I use across work and personal projects. Each skill teaches an AI agent how to handle a specific task, from rewriting copy to running a structured design-to-ship workflow.
Install with Agent Kit from npm. No global install required.
Development harness
A spec-driven pipeline for building software
Twelve skills chain into a spec-driven pipeline across design, change control, plan, build, and ship. You maintain docs/technical/ and approve impact reports; the agent implements, tests, and ships within frozen scope. Each step writes a reviewable artifact and hands off to the next.
This sits alongside the agent harness your editor already provides (Cursor, Claude Code, and the rest). Those tools supply the loop, tools, and context. These skills supply the engineering workflow wrapped around them.
Greenfield: design → plan → build → ship
brainstorming
docs/designs/
technical-documentation
docs/technical/
planning
docs/plans/
project-setup
scaffold repo
executing-plans
task batches + checkpoints
shipping
PR
Brownfield: doc change → delta → ship
refining-docs
docs/technical/
conformance-check
drift audit
reconciling-changes
impact report
planning
delta plan
executing-plans
surgical edits
shipping
PR
Change control
conformance-check + reconciling-changes
What makes brownfield work safe. Audit spec–code alignment, classify doc changes, map impact, and define frozen scope before any code changes. The agent stops for your approval on the impact report.
Build loops
test-driven-development + end-to-end-testing
TDD is the inner loop (red → green → refactor with node:test, tests named with REQ-IDs). E2E is the outer gate: if the running app fails, control returns to the build loop.
Documentation layout
Living documents in docs/technical/ are always current. Dated files in docs/designs/ and docs/plans/ record point-in-time decisions.
text
docs/
designs/ # brainstorming: point-in-time, dated
technical/ # technical-documentation: living, canonical
requirements.md # REQ-001, REQ-001-AC1, acceptance criteria
CHANGELOG.md # REQ-level doc change history
plans/ # planning, impact, conformance: point-in-time
2026-06-29-feature.md # implementation plan
2026-07-01-export-impact.md # impact report
2026-07-02-conformance-report.md # drift auditRead the full story in A development harness for building software with AI agents.
Browse
Skills by category
Seventeen skills organised into seven categories. Each category serves a distinct purpose in the development workflow.
Design & Documentation
Shape ideas, write technical docs, and keep them accurate
Brainstorming
Think through an idea with a trusted colleague before building. Writes design specs.
Technical Documentation
Turn an approved design into production-ready technical docs an AI agent can build from.
Refining Docs
Refine and tune an existing technical doc set: gap analysis, consistency, diagrams, decision tuning.
UI & Component Design
Design systems, component development, and UX best practices
Design System
Define design foundation: brand identity, design tokens, colour palette, typography, spacing, and component standards.
UI/UX Best Practices
Review and guide UI/UX work for clean, professional, low-noise interfaces. Checks hierarchy, whitespace, responsiveness, and accessibility.
Component Library
Set up component library showcase infrastructure with Storybook 8 or custom gallery. Creates foundation for component-driven development.
Component Development
Build UI components in isolation following CDD best practices. Creates components with stories, variants, interaction tests, and documentation.
Component Testing
Add comprehensive testing to components: interaction tests, visual regression, accessibility audits, and unit tests.
Change Control
Audit alignment, analyse impact, and define frozen scope
Planning & Setup
Turn specs into executable tasks and scaffold projects
Build & Testing
Execute plans with TDD and verify the running app
Executing Plans
Execute a written implementation plan task by task, driving the TDD loop and e2e verification.
Test-Driven Development
The core build loop: red-green-refactor with node:test until every requirement passes.
End-to-End Testing
Verify the running app against requirements; retrigger the build loop on failure.
Shipping
Branch hygiene, commits, changelog, and pull requests
Workflow
Development harness pipeline
Design
Shape the idea, write technical docs, and keep them accurate as decisions land.
docs/designs/ → docs/technical/
Step 1 · Design
Brainstorming
Think through an idea with a trusted colleague before building. Writes design specs.
// workflow;Step 2 · Design
Technical Documentation
Turn an approved design into production-ready technical docs an AI agent can build from.
// workflow;Design support
Refining Docs
Refine and tune an existing technical doc set: gap analysis, consistency, diagrams, decision tuning.
// workflow;Change control
Audit spec–code alignment, analyse doc deltas, and define frozen scope before brownfield edits.
docs/plans/ (impact + conformance reports)
Step 3 · Change control
Conformance Check
Audit docs/technical/ against the codebase: missing tests, orphan tests, locked stack violations, contract drift.
// workflow;Step 4 · Change control
Reconciling Changes
Analyse what changed in docs/technical/, map impact to code and tests, and define frozen scope before edits.
// workflow;Plan
Turn approved docs into bite-sized, test-driven tasks an agent can execute.
docs/plans/
Build
Scaffold the repo, execute the plan with TDD, and verify the running app.
src/ + test/
Ship
Branch hygiene, conventional commits, changelog, and a pull request.
PR
Verification loops
Used inside executing-plans. The inner loop builds; the outer gate proves the running app works.
Inner loop
Test-Driven Development
The core build loop: red-green-refactor with node:test until every requirement passes.
// workflow;Outer gate
End-to-End Testing
Verify the running app against requirements; retrigger the build loop on failure.
// workflow;Element loop
Component Development
Build UI components in isolation following CDD best practices. Creates components with stories, variants, interaction tests, and documentation.
// workflow;Element verification
Component Testing
Add comprehensive testing to components: interaction tests, visual regression, accessibility audits, and unit tests.
// workflow;sh
npx @levi-putna/agent-kit@latest add levi-putna/skills --globalOr install individually. See each skill page for targeted commands.
Catalogue
All skills
General Writing
Rewrite or draft text in clear, natural language. Strips AI clichés, filler, and overly formal phrasing.
Mermaid Diagrams
Create professional Mermaid diagrams with consistent theming, semantic colours, and layout rules.
Brainstorming
Think through an idea with a trusted colleague before building. Writes design specs.
Technical Documentation
Turn an approved design into production-ready technical docs an AI agent can build from.
Refining Docs
Refine and tune an existing technical doc set: gap analysis, consistency, diagrams, decision tuning.
Conformance Check
Audit docs/technical/ against the codebase: missing tests, orphan tests, locked stack violations, contract drift.
Reconciling Changes
Analyse what changed in docs/technical/, map impact to code and tests, and define frozen scope before edits.
Planning
Turn specs into executable TDD task lists: greenfield (full scope) or delta (changed REQs only).
Project Setup
Scaffold a new project from technical docs: structure, node:test, tooling, and CI.
Executing Plans
Execute a written implementation plan task by task, driving the TDD loop and e2e verification.
Test-Driven Development
The core build loop: red-green-refactor with node:test until every requirement passes.
End-to-End Testing
Verify the running app against requirements; retrigger the build loop on failure.
Shipping
Ship verified work: branch hygiene, conventional commits, changelog, and a pull request.
Design System
Define design foundation: brand identity, design tokens, colour palette, typography, spacing, and component standards.
UI/UX Best Practices
Review and guide UI/UX work for clean, professional, low-noise interfaces. Checks hierarchy, whitespace, responsiveness, and accessibility.
Component Library
Set up component library showcase infrastructure with Storybook 8 or custom gallery. Creates foundation for component-driven development.
Component Development
Build UI components in isolation following CDD best practices. Creates components with stories, variants, interaction tests, and documentation.
Component Testing
Add comprehensive testing to components: interaction tests, visual regression, accessibility audits, and unit tests.
Source
Browse and install
sh
npx @levi-putna/agent-kit@latest list levi-putna/skills