gsd-plan-review-convergence

Automate cross-AI plan review loops until HIGH-severity concerns are resolved.

1|Updated Sep 6, 2015
One-click install
npx skills add https://github.com/deerawan/dotfiles --skill gsd-plan-review-convergence-deerawan
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gsd-plan-review-convergence
Source: https://github.com/deerawan/dotfiles/tree/main/claude/skills/gsd-plan-review-convergence
Command: npx skills add https://github.com/deerawan/dotfiles --skill gsd-plan-review-convergence-deerawan

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It prevents brittle plans by repeatedly reviewing a plan with external AI reviewers and forcing re-planning when only the highest-severity (“HIGH”) concerns remain.

Core Features & Use Cases

  • Cross-AI revision loop: runs plan review with selectable external AI CLIs (Codex, Gemini, Claude, OpenCode) and optionally local OpenAI-compatible servers (Ollama, LM Studio, llama.cpp).
  • HIGH-concern gate: continues the replan → re-review cycle until no HIGH concerns remain or a max cycle limit is hit.
  • Workflow orchestration: acts as an outer revision gate around existing Skills, delegating actual plan generation and review to them while the orchestrator controls loop control and escalation.

Quick Start

Run the skill with the phase number for the plan you want to converge, enabling the external reviewers you prefer and allowing it to iterate until HIGH concerns are resolved.

Frequently Asked Questions about gsd-plan-review-convergence

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

FAQPage Schema
How do I automate plan review and replanning until AI finds no high severity issues?

You automate a cross-AI plan convergence loop by running plan reviews with external AI CLIs and triggering replanning when HIGH-severity concerns are detected. The cycle continues until no HIGH concerns remain or a max cycle limit is hit.

Can I use external AI reviewers like Codex or Gemini for iterative workflow planning?

Yes, you can select external AI reviewers like Codex, Gemini, Claude, or OpenCode for iterative workflow planning. You can also optionally use local OpenAI-compatible servers such as Ollama, LM Studio, or llama.cpp.

What is a cross-AI revision loop for quality gates?

A cross-AI revision loop for quality gates is an automated cycle that repeatedly reviews a generated plan using external reviewer CLIs and forces replanning when HIGH-severity concerns are detected, ensuring plans are not brittle.

How do I stop brittle plans from passing review thresholds?

You prevent brittle plans by enforcing a HIGH-concern gate that continues the replan and re-review cycle until no HIGH severity concerns remain. This loop control acts as an outer revision gate around existing plan generation.

Does this plan review convergence loop support max-cycle constraints?

Yes, the plan review convergence loop supports max-cycle constraints to prevent infinite loops. It also enforces a feature gate, supports reviewer selection, and preserves revision and escalation workflow safety checks.

Do I need local AI servers to run multi-model agent orchestration for plan review?

No, you do not need local AI servers. Multi-model agent orchestration supports external reviewer CLIs like Codex and Claude, but you can optionally use local OpenAI-compatible servers like Ollama or LM Studio.