zmackie-skillz

skillz

A Claude Code plugin with skills for harness engineering — configuring, optimizing, and verifying coding agent workflows — plus creativity, security, and general dev workflow skills.

Works with Claude Code, Codex, Cursor, Windsurf, Copilot, and custom agent systems.

Install

/plugin marketplace add zmackie/skillz
/plugin install skillz@zmackie-skillz

Which skill should I use?

I want to...	Start with	Then consider
Set up or improve my agent workflow	harness-audit	The skills it recommends based on findings
Write CLAUDE.md / AGENTS.md / .cursorrules	write-agent-instructions	verification-harness to enforce them
Fix an agent losing context on long tasks	context-architecture	multi-agent-design if decomposition helps
Make my codebase easier for agents to navigate	agent-legibility	write-agent-instructions for the instruction side
Add hooks, linters, or CI checks for agent quality	verification-harness	harness-audit to identify what's missing
Design multi-agent pipelines	multi-agent-design	context-architecture for isolation patterns
Brainstorm creative or novel ideas	cross-domain-ideation	idea-garden to prototype the best ones
Prototype ideas autonomously in the background	idea-garden	vibe-research to write up results
Write a design doc, RFC, or research report	vibe-research	--
Find security vulnerabilities in code	vuln-research	code-review for lighter review
Review code for quality	code-review	vuln-research for deeper security focus
Generate a commit message	commit-message	--
Set up or work on a Python project	python-uv	verification-harness for pytest/mypy hooks
Integrate LLM APIs	openrouter-api	python-uv for Python project setup

Skill Map

Harness Engineering

Start with harness-audit to assess your current setup — it diagnoses issues across five areas and points you to the right specialized skill.

harness-audit (start here)
  |
  |-- write-agent-instructions    Write/improve CLAUDE.md, AGENTS.md, etc.
  |-- verification-harness        Set up hooks, linters, test runners, CI checks
  |-- agent-legibility            Optimize codebase for agent navigation
  |-- context-architecture        Design sub-agents, resets, progressive disclosure
  |-- multi-agent-design          Planner/generator/evaluator patterns

Creativity & Research

Three skills that form a generate -> prototype -> write-up pipeline:

Skill	What it does
cross-domain-ideation	Generate novel ideas using analogical mapping from distant domains
idea-garden	Autonomously prototype ideas from a backlog (vapor -> fermenting -> liquid -> frozen)
vibe-research	Produce polished written artifacts through multi-agent review loops

Security

Skill	What it does
vuln-research	LLM-assisted vulnerability research: threat modeling, slice-based auditing, adversarial prompting, PoC validation

General Dev

Skill	What it does
code-review	Structured code review (security, performance, correctness, readability)
commit-message	Generate conventional-commit messages from staged diffs
python-uv	Python project conventions (uv, click, pytest, type hints)
openrouter-api	LLM integration patterns via OpenRouter

Usage

Invoke skills as /skillz:<skill-name>:

/skillz:harness-audit
/skillz:write-agent-instructions
/skillz:code-review

Most skills also auto-trigger when Claude detects a relevant task.

Adding a Skill

Create skills/<name>/SKILL.md with YAML frontmatter:

---
name: my-skill
category: harness-engineering  # or: creativity, security, general-dev
description: What it does and when to use it
---

Instructions for Claude when this skill is active.

See meta-skill.md for writing guidelines and AGENTS.md for plugin structure.