Does Plan Forge save previous versions of plans?

Yes, it maintains versioned markdown files in your local directory, allowing you to track how the plan evolved through each review round.

What are the Alpha and Beta review lenses?

Alpha (Forge Inspector) focuses on correctness and soundness, while Beta (Forge Optimizer) focuses on simplification and preventing over-engineering.

How does Plan Forge differ from /plan-review?

While /plan-review is a one-shot review of an existing plan, Plan Forge creates the plan from scratch and runs up to three rounds of dual-agent review, consolidation, and revision.

When is it best to use Plan Forge?

Use it for complex, multi-file features or architectural changes where a mistake in planning would be more expensive than the time spent in review.

Can I skip the creation phase and just review an existing file?

Yes, by using the --review-only flag followed by the path to your plan, you can trigger the forge cycles on pre-existing documentation.

Plan Forge

Name: Plan Forge
Author: ahrav

byahrav

•

開発者ツール

Iteratively creates and stress-tests implementation plans through multi-agent review cycles to catch architectural blind spots before coding begins.

Plan Forge transforms the planning process into a rigorous metallurgical workflow: implementation plans are heated through review, hammered via revision, and quenched into a finalized artifact. Unlike standard one-shot planning, it employs dual-agent reviews—Alpha (Inspector) for correctness and Beta (Optimizer) for simplicity—to stress-test logic against the actual codebase. By identifying performance pitfalls, race conditions, and unnecessary complexity while changes are still 'cheap,' it ensures that the implementation phase is efficient, sound, and free of architectural surprises.

主な機能

01Dual-agent perspectives focusing on Correctness/Soundness vs. Simplification/Pragmatism

02Grounds all findings in codebase reality using integrated Glob, Grep, and Read tools

031 GitHub stars

04Iterative multi-round review cycles (1-3 rounds) for maximum plan stability

05Automated finding consolidation with impact and confidence scoring

06Versioned plan storage (~/.claude/plans/) for historical tracking and diffing

ユースケース

01Validating non-obvious architectural decisions before writing a single line of code

02Planning multi-step features that touch critical code paths or sensitive business logic

03Simplifying complex implementation ideas to avoid YAGNI violations and technical debt

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ahrav/gossip-rs plan-forge

For use in Claude.ai and ChatGPT