What is the Orchestrator skill for Claude Code?

It is a specialized capability that enforces a disciplined workflow for complex tasks, utilizing structured planning, subagent delegation, and persistent documentation to ensure high-quality outcomes.

What is the self-improvement loop?

Whenever a user provides a correction or a bug is found, the skill captures the 'pattern' and 'rule' in a lessons file, ensuring Claude internalizes the feedback and avoids repeating the mistake in the future.

When should I trigger the Orchestrator?

You should use it for non-trivial tasks involving three or more steps, architectural decisions, or when you explicitly need a methodical, autonomous approach by saying 'orchestrate' or 'plan this out'.

How does it track progress across sessions?

The skill maintains local files in .claude/output/orchestrator/, specifically todo.md for task status and lessons.md for institutional memory, which are read during the session bootstrap phase.

How do subagents work within this skill?

The orchestrator offloads research, exploration, and focused execution to subagents, allowing the main Claude instance to focus on high-level strategy while keeping the primary context window clean.

Structured Workflow Orchestrator

Name: Structured Workflow Orchestrator
Author: flosrn

byflosrn

0•

生産性とワークフロー

Orchestrates complex development tasks using a plan-first methodology, subagent delegation, and automated verification gates.

The Structured Workflow Orchestrator transforms Claude into a high-level staff engineer by enforcing a disciplined, multi-phase methodology for non-trivial tasks. It prioritizes architectural planning and progress tracking via local Markdown files, ensuring that context is maintained across sessions. By delegating focused research and execution to subagents, it preserves the main context window for high-level decision-making. The skill includes a unique self-improvement loop that captures lessons from mistakes, alongside rigorous verification gates to ensure that no code is marked complete until its correctness is proven.

主な機能

01Plan-first methodology with mandatory progress tracking in todo.md

02Rigorous verification gates to ensure code quality and staff-engineer level approval

03Automated subagent delegation for research and parallel task execution

040 GitHub stars

05Persistent self-improvement loop using lessons.md to avoid repeating past errors

06Session bootstrapping to maintain institutional memory across work sessions

ユースケース

01Managing large-scale feature implementations through parallel subagent coordination

02Executing complex architectural refactors that require multi-step planning and dependency tracking

03Autonomous bug fixing that involves deep root cause analysis and log investigation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add flosrn/.claude orchestrator

For use in Claude.ai and ChatGPT