What is the 'plans are prompts' principle?

It means the generated PLAN.md files are not just documentation; they are structured instructions designed to be read and executed directly by Claude as a system prompt.

Can I use this for specific platforms like iOS or macOS?

Yes, the skill includes domain expertise modules that load specific principles and references for platforms like macOS, iOS, Unity, and specialized UI design.

What happens if Claude encounters a bug during plan execution?

The skill uses five embedded deviation rules that allow Claude to automatically fix bugs and blockers while documenting the changes in a summary file for the user.

How does this skill prevent AI quality degradation?

It enforces aggressive atomicity, splitting phases into small plans of 2-3 tasks to ensure work stays within the first 50% of the context window where Claude's quality is highest.

Agentic Project Planner

Name: Agentic Project Planner
Author: Eric-Ness

byEric-Ness

Productivity & Workflow

Generates hierarchical, Claude-executable project plans optimized for solo agentic development and context management.

About

The Agentic Project Planner skill transforms high-level project visions into granular, executable roadmaps specifically designed for the interaction between a solo developer and Claude. Unlike traditional enterprise documentation, this skill produces 'plans as prompts'—atomic, verifiable units of work that prevent AI quality degradation by keeping context usage low. It automates the creation of briefs, roadmaps, and phase plans, incorporating standardized deviation rules and human checkpoints to ensure development flow remains uninterrupted while maintaining high technical standards.

Key Features

Human-in-the-loop checkpoints for verification and critical decision gates
Atomic task splitting to maintain peak AI output quality (0-50% context usage)
Hierarchical planning from high-level vision to specific execution files
Automated deviation rules for bug fixes and blockers without breaking flow
Intelligent domain expertise loading for macOS, iOS, Unity, and web development
0 GitHub stars

Use Cases

Breaking down a complex feature request into a series of independently verifiable atomic plans
Automating CLI-driven workflows where Claude handles execution and the user provides verification
Managing context-heavy development tasks by creating structured handoffs and milestones

About

Key Features

Human-in-the-loop checkpoints for verification and critical decision gates
Atomic task splitting to maintain peak AI output quality (0-50% context usage)
Hierarchical planning from high-level vision to specific execution files
Automated deviation rules for bug fixes and blockers without breaking flow
Intelligent domain expertise loading for macOS, iOS, Unity, and web development
0 GitHub stars

Use Cases

Breaking down a complex feature request into a series of independently verifiable atomic plans
Automating CLI-driven workflows where Claude handles execution and the user provides verification
Managing context-heavy development tasks by creating structured handoffs and milestones