
Wrangler
Add wrangler to Claude Code to design and run Planner → Generator → Evaluator multi-agent harnesses for complex, long-running tasks.
Overview
wrangler is a plugin marketplace for the Build phase that helps you design and run Planner → Generator → Evaluator multi-agent harnesses in Claude Code.
What is this marketplace?
- Official Anthropic marketplace schema entry with development category and GitHub homepage
- Documents a three-role pipeline: Planner → Generator → Evaluator
- Targets complex, long-running AI tasks rather than single-shot prompts
- Single wrangler plugin sourced from ./; marketplace title: multi-agent harness engineering for Claude Code
- 1 plugin: wrangler
- 3 documented agent roles in pipeline
- Marketplace uses anthropic.com/claude-code/marketplace.schema.json
What problem does it solve?
Long, complex agent jobs fail when one model thread plans, codes, and judges quality without a durable multi-role harness.
Who is it for?
Advanced solo builders orchestrating multi-step agent work in Claude Code who need explicit planner, generator, and evaluator roles.
Skip if: Beginners doing short edits or teams that only need linting, CRUD codegen, or a single MCP tool.
What do I get? / Deliverables
After installing wrangler, you can stand up a Planner → Generator → Evaluator harness in Claude Code and rerun it on complex, long-running work with clearer evaluation gates.
- Enabled wrangler development plugin
- Documented Planner → Generator → Evaluator harness pattern
- Repeatable long-running task orchestration inside Claude Code
Plugins in this marketplace
1 plugin — install individually after you add the marketplace.
Recommended Marketplaces
Journey fit
Harness engineering is a Build concern first because you assemble agent roles and orchestration before you rely on eval loops at ship time. agent-tooling is the canonical home for multi-agent role design, orchestration patterns, and Claude Code plugin harnesses.
How it compares
Multi-agent harness engineering plugin for Claude Code, not a curated skill pack or executive work-management bundle.
Common Questions / FAQ
Who is Wrangler for?
Claude Code users running complex, long-running agent tasks who want a structured Planner → Generator → Evaluator harness.
When should I use Wrangler?
Use it when a task spans many steps, needs planned decomposition, and requires evaluator gates before you trust the output.
How do I add Wrangler to my agent?
Add the panicgit/Wrangler marketplace in Claude Code, enable the Wrangler plugin from ./, and follow the repo harness patterns linked from the marketplace homepage.