nklisch-skills

nklisch/skills

A software development workflow suite for Claude Code — from ideation through implementation, refactoring, testing, and release.

Available as a Claude Code plugin and via skilltap.

Install

As a Plugin (Claude Code)

# Step 1: Add the marketplace
/plugin marketplace add nklisch/skills

# Step 2: Install plugins
/plugin install workflow@nklisch-skills
/plugin install skill-authoring@nklisch-skills

Two plugins available:

• workflow — 21 skills for the full development pipeline
• skill-authoring — 3 skills for creating, evaluating, and refining agent skills

Via Skilltap

# Add this tap
skilltap tap add nklisch https://github.com/nklisch/skills

# Install individual skills
skilltap install design
skilltap install implement
skilltap install research

Workflow Skills

These skills form an iterative pipeline. See docs/workflow-guide.md for the full usage guide.

Pipeline

ideate → roadmap                    ← project start
    ↓
design → implement/implement-orchestrator ← core phases (from roadmap)
    ↓
refactor-design → implement → extract-patterns  ← every 2-4 phases
    ↓
bold-refactor / perf-design → implement         ← targeted improvements
    ↓
expand                              ← major scope changes (updates foundation docs)
    ↓
stylistic/structural-refactor-creator         ← every 3-5 phases
    ↓
security-review → design → implement           ← security audit pass
    ↓
feature → design → implement        ← quick one-offs near the end
    ↓
release                              ← ship it

Skill	What it does
ideate	Interactive project definition workshop. Produces foundation docs (VISION.md, SPEC.md, ARCHITECTURE.md). Run once at project start.
roadmap	Generate an agent-optimized project roadmap from foundation docs. Decomposes into phases with dependencies and test checkpoints.
expand	Expand project scope — new subsystem, major capability, architectural shift. Updates foundation docs and roadmap.
feature	Scope a quick extension or one-off outside the core roadmap. Produces a feature brief that design consumes.
design	Produce detailed implementation units with exact interfaces, types, and acceptance criteria.
implement	Write code from a design or plan. Single Sonnet agent. Best for <20 files.
implement-orchestrator	Opus orchestrator that spawns Sonnet agents. For 20+ files or independent subsystems.
research	Investigate libraries, APIs, and SDKs. Produces a research doc and an auto-loading reference skill.
refactor-design	Find duplication and missing abstractions. Produces a refactor plan that implement executes.
bold-refactor	Find beautiful code abstractions and cross-cutting simplifications using conceptual lenses. Produces an implement-ready design.
extract-patterns	Document reusable code patterns for consistency across future work.
stylistic-refactor-creator	Interview-based. Generates a project-specific stylistic-refactor skill.
structural-refactor-creator	Interview-based. Generates a project-specific structural-refactor skill.
perf-design	Profile performance bottlenecks and design optimized solutions. Follows a strict optimization hierarchy.
security-review	Comprehensive security audit. Discovers stack, user selects focus domains, produces scored report with severity-classified findings.
cruft-cleaner	Sweep for AI-accumulated cruft (dead code, stale comments, defensive bloat). Triages by confidence, parallel cleanup.
test-quality	Spec-driven test gap analysis. Derives tests from contracts, not code. Writes the tests.
e2e-test-design	Design golden-path and adversarial e2e test suites. Interactive.
update-documentation	Align all docs to code after changes.
release	Draft changelog, confirm with user, run the project's release mechanism.

Skill Authoring

Skill	What it does
write-tool-skill	Create distributable reference skills for your project's tool, CLI, MCP server, or library — for others to install.
skill-idea-refiner	Refine a rough skill idea into a well-designed skill. Guides through ideation, scoping, and scaffolding.
skill-evaluator	Evaluate skills against type-specific quality rubrics. Produces scored reports with improvements.
tool-evaluator	Self-evaluate agent tool usage in the current conversation. Produces a report with recommendations for tool authors.

Principles (auto-load)

Auto-load during design and implementation to enforce standards.