What effort model does the skill use?

By default, it uses a T-shirt sizing model (e.g., S, M, L, XL) as defined in the project's effort model configuration.

What happens if the planning phase requires more information?

If clarification is needed, the skill creates a question file in the .swarm/pending-questions/ directory and sets the story status to 'awaiting_input' instead of 'planned'.

How does it ensure all requirements are met?

The skill includes a 'Validate Coverage' step that ensures every Acceptance Criterion (AC) is mapped to at least one specific task through an ac_coverage field.

What is the primary purpose of the 'plan' skill?

The 'plan' skill transitions a development story from the 'ideating' status to 'planned' by creating a technical approach, breaking it into tasks, and assigning them to agents.

How does it assign agents to tasks?

It reads the agent roles and Ways of Working (WoW) from your project's .swarm/config.md file to match tasks with the most appropriate agent roles.

Recursive AI Project Planner

Name: Recursive AI Project Planner
Author: hgeldenhuys

byhgeldenhuys

0•

生产力与工作流

Automates the transition from ideation to execution by creating task breakdowns, assigning agents, and estimating effort for AI swarms.

The Recursive AI Project Planner is a specialized skill for Claude Code designed to handle the complex 'planning' phase of software development. It systematically analyzes project backlogs, validates acceptance criteria, and designs a technical approach before breaking a story down into granular, actionable tasks. By integrating with the .swarm configuration, it intelligently assigns specialized agents to specific tasks, maps out dependencies, and applies effort estimations using standardized models like T-shirt sizing, ensuring a seamless transition from a high-level idea to a project-ready state.

主要功能

01Automated story status transitions within the swarm workflow

02Automated task breakdown with YAML-formatted outputs

03Comprehensive acceptance criteria (AC) coverage validation

040 GitHub stars

05Intelligent agent assignment based on role and expertise

06Dependency mapping to prevent circular task logic

使用场景

01Converting high-level feature descriptions into structured, agent-ready backlogs

02Ensuring architectural consistency and effort estimation accuracy in AI-driven projects

03Orchestrating complex multi-agent software engineering workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add hgeldenhuys/recursive-ai plan

For use in Claude.ai and ChatGPT

Download Skill