autonomous-agent-harness
Set up autonomous coding agent projects with long-running harnesses using Archon MCP for state management. Creates complete project scaffolds with initializer/coding agent prompts, feature tracking, session handoffs, security configuration, and browser testing integration. Based on Anthropic's effective harnesses guide. Use when building autonomous coding agents, long-running AI workflows, or multi-session development projects.
npx skills add majiayu000/claude-skill-registry --skill autonomous-agent-harness --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Creates fully-configured autonomous coding agent projects capable of multi-session operation. Uses Archon MCP for state management across projects, tasks, and documents. Generates a complete scaffold including initializer/coding prompts, session handoffs, feature tracking, security configuration, and browser testing integration. Provides an agent pipeline with specialized roles (wizard, initializer, coder, tester, reviewer) and multiple execution modes (terminal, background, SDK).
How it works
- Guided quick start prompts define commands like
/harness-setup,/harness-init,/harness-next, and/harness-statusto drive setup and progress. - Architecture maps a multi-agent pipeline: harness-wizard creates initial tasks via /harness-setup → @harness-wizard → /harness-init, then /harness-next drives features through @harness-coder with parallel testing via @harness-tester and review via @harness-reviewer.
- Archon MCP handles state management for projects, tasks, and documents; initialization creates a project scaffold and configuration files.
- Prompts are provided for initializer and coding continuation sessions to instruct agents on task creation, handoffs, and testing requirements.
- Generates environment and project structure, including .archon_project.json, .claude_settings.json, app_spec.txt, init.sh, and prompts/ files.
When to use it
- Use when building autonomous coding agents, long-running AI workflows, or multi-session development projects.
- Trigger initial setup with /harness-setup and start the first session with /harness-init; continue work with /harness-next and monitor status with /harness-status.
What it can touch
- Generates and edits project scaffolding under a new directory named after the project; creates and writes several configuration and prompt files:
- ".archon_project.json"
- ".claude_settings.json"
- "app_spec.txt"
- "init.sh"
- "claude-progress.txt"
- "features.json"
- files under "prompts/": initializer_prompt.md, coding_prompt.md
- directory "src/", "tests/", "docs/"
- Interacts with Archon MCP servers for project, task, and document management, and can utilize Playwright MCP for browser testing.
Caveats
- Behavior relies on Archon MCP setup and access; misconfiguration could impede state management.
- The skill defines a workflow with multiple agents and steps; actual outcomes depend on execution and test results within the MCP environment.
- Filesystem and command execution policies are configurable via the generated claude_settings.json; certain bash commands are denied for safety.
# Autonomous Coding Agent Harness Setup Create fully-configured autonomous coding agent projects that can work across multiple sessions with proper state management, handoffs, and testing. Uses **Archon MCP** for project/task tracking, enabling persistent state management and context preservation. ## 🚀 Quick Start Use these prompts to interact with the harness system: | Command | Description | |---------|-------------| | `/harness-setup` | Launch full setup wizard | | `/harness-quick` | Quick setup with smart defaults | | `/harness-init` | Initialize project (first session) | | `/harness-next` | Start next coding session | | `/harness-status` | Check project status | | `/harness-resume` | Resume existing project | --- ## Architecture Overview ``` ┌─────────────────────────────────────────────────────────────────────┐ │ MULTI-AGENT PIPELINE │ │ │ │ /harness-setup → @harness-wizard │ │ │ │ │ ▼ │ │ /harness-init → @harness-initializer │ │ │ │ │ ▼ │ │ ┌─────────────────────────────────────────────────────────────┐ │ │ │ /harness-next → @harness-coder │ │ │ │ │ │ │ │ │ ├──► @harness-tester (parallel) │ │ │ │ │ │ │ │ │ ├──► @harness-reviewer (before completion) │ │ │ │ │ │ │ │ │ ▼ │ │ │
- 🚀 Quick Start
- Architecture Overview
- Agent Pipeline
- Features
- Project Setup Questionnaire
- Phase 1: Project Basics
- Phase 2: Technical Stack
- Phase 3: Agent Configuration
- Phase 4: Testing & Security
- Phase 5: Archon Integration
- Phase 6: Application Specification
- Project Generation Workflow
- Step 1: Create Archon Project
- Step 2: Generate Directory Structure
set -e
echo "Initializing {PROJECT_NAME}..."
Create directories
mkdir -p src tests docs
Initialize git if not already
if [ ! -d ".git" ]; then
git init
echo "node_modules/" >> .gitignore
echo ".env" >> .gitignore
echo "__pycache__/" >> .gitignoreWhat does the autonomous-agent-harness skill do?
Set up autonomous coding agent projects with long-running harnesses using Archon MCP for state management. Creates complete project scaffolds with initializer/coding agent prompts, feature tracking, session handoffs, security configuration, and browser testing integration. Based on Anthropic's effective harnesses guide. Use when building autonomous coding agents, long-running AI workflows, or multi-session development projects.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill autonomous-agent-harness --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.