skills.rest
leojkwan
leojkwanCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

vidux

Enforce plan-first orchestration for AI agents using PLAN.md as the source of truth.

Updated Aug 23, 2026
Search Tags:#git#markdown#hooks#evidence#ai orchestration#plan-first
One-click install
npx skills add https://github.com/leojkwan/vidux --skill vidux
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: vidux
Source: https://github.com/leojkwan/vidux/tree/main
Command: npx skills add https://github.com/leojkwan/vidux --skill vidux

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Enforces plan-first orchestration for AI agents by treating PLAN.md as the single source of truth and guiding every cycle.

Core Features & Use Cases

  • Evidence-driven planning: gather and cite evidence before acting.
  • Plan-to-code discipline: ensure code changes align with PLAN.md checkpoints.
  • Hook-driven enforcement: pre/post tool hooks and session resume to maintain discipline.
  • Fleet compatibility: works with multi-session Claude/Codex fleets and optional external adapters.
  • Stateless loop: reads PLAN.md at each cycle and checkpoints changes back to git.
  • Use case: coordinate long-running AI improvements across days with guaranteed traceability.

Quick Start

Install Vidux, symlink the skill into Claude Code's skills path, then start a session with /vidux to begin.

Frequently Asked Questions about vidux

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

FAQPage Schema
How do I enforce plan-first orchestration for AI coding agents?▼

You can enforce plan-first orchestration by using a markdown-based contract that treats PLAN.md as the single source of truth, guiding every AI agent cycle with structured checkpoints and hook-driven execution.

What is evidence-driven planning for AI agents?▼

Evidence-driven planning requires AI agents to gather and cite evidence before taking action. This ensures code changes align with PLAN.md checkpoints and maintains deterministic execution throughout the workflow.

How do I coordinate long-running multi-session AI coding improvements?▼

You can coordinate long-running AI improvements by applying a stateless loop that reads PLAN.md at each cycle and checkpoints changes back to git. This supports multi-session fleets of Claude Code and Codex agents.

Can I use Claude Code and Codex fleets with a single PLAN.md file?▼

Yes, this orchestration approach works with multi-session Claude Code and Codex fleets. It uses optional external adapters to maintain plan-to-code discipline across all agents referencing the same PLAN.md.

How does the stateless execution loop work with git checkpoints?▼

The stateless loop reads PLAN.md at each cycle, executes a five-step process of Read, Assess, Act, Verify, and Checkpoint, then commits changes back to git. This ensures actions remain aligned with the plan.

What are the limitations of using a markdown file for AI agent orchestration?▼

Relying on a single PLAN.md file requires strict pre and post tool hooks to maintain discipline. If session resume fails or agents bypass these enforcement hooks, plan-to-code alignment may break across multi-session fleets.

Related Skills

Explore More Software Engineering Skills→
8Dionysus8Dionysus
4

aoa-source-of-truth-check

Identify and resolve documentation ambiguity to establish a single source of truth.

Community
Intermediate
vanhoangkhavanhoangkha

prowler-commit

Generate Git commit messages following the Conventional Commits specification.

Community
Intermediate
thmtzthmtz
5

add-codex

Replace Claude Agent SDK with Codex for planning, tool orchestration, and session resume.

Community
Intermediate
The13thNodeThe13thNode
3

head-of-growth

Conduct weekly growth audits and design experiments for startup metrics.

Community
Advanced