By sliday
Harness Engineering for AI coding agents — CAR framework, security guardrails, quality gates, context monitoring, auto-formatters, trace logging, checkpoints, MCP validation
npx claudepluginhub sliday/harnBuild or rebuild AGENTS.md with progressive disclosure
Audit your project's harness against the CAR framework
Generate resume artifacts — LEARNED.md, CHECKPOINT.json — for long-running sessions
Add context monitoring and backpressure hooks to manage agent attention budget
Generate or update PostToolUse auto-formatter hook
Generate or update quality gate Stop hook
Generate or update security guardrail hooks
Scaffold a complete harness for the current project — AGENTS.md, hooks, guardrails
Generate MCP server health check hook
Generate PostToolUse trace logging hook for agent observability
Audit project harness against the CAR framework. Use when: /harn:analyze, 'audit harness', 'check CAR', 'harness health', 'how good is my harness'
Scaffold resume artifacts (LEARNED.md, CHECKPOINT.json) for long-running sessions. Triggers: /harn:checkpoint, 'resume artifacts', 'checkpoint', 'session recovery', 'LEARNED.md'
Generate context monitoring and backpressure hooks. Triggers: /harn:context, 'context monitoring', 'backpressure', 'context budget', 'attention budget', 'context rot'
Scaffold a complete harness for the current project — AGENTS.md, hooks, guardrails. Use when starting a new project or hardening an existing one. Triggers: /harn:init, 'set up harness', 'scaffold guardrails', 'create AGENTS.md'
Generate MCP server health check for SessionStart. Triggers: /harn:mcp-check, 'MCP health', 'check MCP servers', 'validate MCP'
Generate or update the quality gate Stop hook. Use when: /harn:quality-gate, 'set up type checking', 'add quality gate', 'stop hook'
Generate or update security guardrail hooks. Use when: /harn:guard, 'update security rules', 'block dangerous commands', 'add guardrail'
Generate PostToolUse trace logging hook for observability. Triggers: /harn:trace, 'trace logging', 'observable hooks', 'agent trace', 'JSONL logging'
Build or rebuild AGENTS.md with progressive disclosure. Use when: /harn:agents-md, 'create AGENTS.md', 'update agent instructions', 'too many instructions'
Generate PostToolUse auto-formatter hook. Triggers: /harn:format, 'auto format', 'formatter hook', 'PostToolUse formatter'
Deterministic guardrails, quality gates, and progressive disclosure for AI coding agents.
The underlying AI model matters less than the system built around it. A well-harnessed agent produces consistent, safe, high-quality output regardless of the model version. Read the foundations.
claude plugins marketplace add sliday/claude-plugins
claude plugins install harn
Then in any project:
/harn:init # scaffold your harness
| Command | Description |
|---|---|
/harn:init | Scaffold a complete harness — AGENTS.md, hooks, guardrails |
/harn:guard | Generate or update security guardrail hooks (PreToolUse) |
/harn:gate | Generate or update quality gate Stop hook |
/harn:analyze | Audit your project against the CAR framework |
/harn:agents-md | Build or rebuild AGENTS.md with progressive disclosure |
/harn:format | Generate PostToolUse auto-formatter hook |
/harn:trace | Generate PostToolUse trace logging hook for observability |
/harn:checkpoint | Generate resume artifacts (LEARNED.md, CHECKPOINT.json) |
/harn:mcp-check | Generate MCP server health check hook |
/harn:context | Monitor context window usage and apply backpressure |
/harn:init CreatesAGENTS.md — Agent behavior spec with progressive disclosure.claude/settings.json — Hook configuration.claude/hooks/security-guard.sh — PreToolUse guardrail.claude/hooks/quality-gate.sh — Stop hook for type/lint/test checks.claude/hooks/auto-format.sh — PostToolUse formatter (optional)LEARNED.md — Persistent gotchas and decisions logEvery agentic system has three layers:
Run /harn:analyze to score your project across all three. Learn more at harn.app.
jq — JSON processing (used by hooks)python3 >= 3.6 — Script executiontsc, cargo, go vet, ruff)If a hook is blocking legitimate work or the agent is stuck in a loop, see references/escape-hatches.md.
Curated articles on harness engineering — security, context, evals, specs, and tools: harn.app/kb.
MIT
Battle-tested Claude Code plugin for engineering teams — 38 agents, 156 skills, 72 legacy command shims, production-ready hooks, and selective install workflows evolved through continuous real-world use
Uses power tools
Uses Bash, Write, or Edit tools
No model invocation
Executes directly as bash, bypassing the AI model
Complete collection of battle-tested Claude Code configs from an Anthropic hackathon winner - agents, skills, hooks, rules, and legacy command shims evolved over 10+ months of intensive daily use
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Orchestrate multi-agent teams for parallel code review, hypothesis-driven debugging, and coordinated feature development using Claude Code's Agent Teams
Comprehensive toolkit for developing Claude Code plugins. Includes 7 expert skills covering hooks, MCP integration, commands, agents, and best practices. AI-assisted plugin creation and validation.