Verified:GitHub

phase-transition

Automate AI session phase transitions with gated checkpoint evaluations.

Updated Aug 23, 2026

Search Tags:#automation #state-management #reproducibility #ai-workflow #phase-transition #checkpoint-eval #context-cleanup

Ready to run

npx skills add https://github.com/ItsMattG/property-tracker --skill phase-transition

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: phase-transition
Source: https://github.com/ItsMattG/property-tracker/tree/main/.claude/skills/phase-transition
Command: npx skills add https://github.com/ItsMattG/property-tracker --skill phase-transition

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill provides a structured mechanism to move AI sessions through defined phases, ensuring context is kept clean and outputs are reproducible across transitions.

Core Features & Use Cases

Enforces a fixed phase order: RESEARCH → PLAN → IMPLEMENT → REVIEW → VERIFY.
Manages per-phase output destinations and a centralized current-phase state for traceability.
Integrates a checkpoint evaluation step to gate transitions and surface findings when criteria are not met.
Use Case: A research investigation transitions to planning, then implementation, review, and verification with preserved context and auditable logs.

Quick Start

Run the phase manager to initialize a new topic and progress through RESEARCH, PLAN, IMPLEMENT, REVIEW, and VERIFY.

Frequently Asked Questions about phase-transition

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I structure AI session phases to maintain context integrity during multi-step workflows?▼

You can structure AI session phases by enforcing a fixed order—RESEARCH, PLAN, IMPLEMENT, REVIEW, VERIFY—using automated transitions. This maintains context integrity by managing per-phase output destinations and updating a centralized current-phase state for traceability.

What is the best way to ensure reproducibility when transitioning between AI research and implementation phases?▼

The best way to ensure reproducibility during phase transitions is to use a centralized current-phase.json file and eval logs. These provide a reproducible starting point and auditable records of state updates and checkpoint evaluations.

How do checkpoint evaluations work when gating transitions between planning and implementation?▼

Checkpoint evaluations gate transitions by surfacing findings when criteria are not met. Before moving from one phase to another, the system evaluates the outputs, ensuring quality and completeness standards are satisfied before progression.

Do I need any external dependencies to automate state management for AI workflows?▼

No external dependencies are required to automate state management for AI workflows. The system operates independently, relying on internal output file management and a current-phase state file to track progress.

Can I customize the phase order if my workflow does not require a review step?▼

The system enforces a fixed phase order of RESEARCH, PLAN, IMPLEMENT, REVIEW, and VERIFY. It is designed for multi-step workflows requiring this specific sequence, meaning skipping or reordering phases like REVIEW is not supported.