claude-wizard

Name: claude-wizard
Author: vlad-ko

64stars20forks

How to install claude-wizard?

vlad-ko/claude-wizard

$git clone https://github.com/vlad-ko/claude-wizard

Installs into the current project.

Summary

From one disciplined developer to an orchestrated team — without giving up the discipline.

Files · 1

View on GitHub

README.md

1	# Wizard v2: the multi-agent workflow
2
3	From one disciplined developer to an orchestrated team — without giving up the discipline.
4
5	Wizard v1 turned Claude Code from a fast coder into a careful one: it read before writing, tested before implementing, and attacked its own code before committing. That's still the foundation. v2 adds a second gear. When the work is big enough, the careful thread stops being the builder and becomes the orchestrator of a team of specialist agents — an issue maintainer who turns ideas into well-formed issues, an architect who designs, builders who implement in parallel, adversarial critics who try to break it, an independent reviewer that catches what the builders missed, and a review gate that won't let anything merge with an unanswered finding. One thread becomes a pipeline that drives a whole cohort of work to merge-ready at once.
6
7	If v1 was "think before you code," v2 is "design, then delegate, then verify — in parallel, without idling." And the most important shift is for you: you stop being a task-giver and become a conductor.
8
9	## Your role: conductor, not task-giver
10
11	This is the defining change of v2. You are not here to write detailed task breakdowns, spell out implementation steps, or write code. The agents do that. Your job is to keep the flow moving from idea → issue → PR → production — and nothing below that altitude.
12
13	Concretely, the human conductor:
14
15	- Sets direction. You bring the idea ("we need transfer-status tracking") — not the task list. Turning that idea into a structured issue with acceptance criteria is the issue-maintainer agent's first step, not your chore.
16	- Makes the product and judgment calls. When an agent hits an ambiguous requirement, a tradeoff, or a "which behavior is correct?" fork, that's yours to answer. The ensemble is deliberately good at building the thing right and deliberately defers deciding what's right to you.
17	- Unblocks. When the orchestrator pings that it's gated on a decision, a credential, or an external dependency, you clear it — then it resumes the whole cohort.
18	- Merges. The orchestrator drives every PR to merge-ready and declares it exactly once; you do the final merge. v2 never auto-merges. The merge button is the one piece of the cycle that stays human.
19
20	Everything between "idea" and "merge button" — issue authoring, design, building, reviewing, answering review findings — is delegated. You conduct; the ensemble plays.
21
22	## What it is
23
24	`/wizard` is a Claude Code [skill](https://docs.anthropic.com/en/docs/claude-code/skills) — a markdown playbook that changes how Claude operates for the duration of a task. v2 runs in two modes:
25
26	- Direct mode — a single thread runs the whole 8-phase lifecycle itself (this is v1's behavior, kept intact).
27	- Delegated mode — the thread you're talking to becomes an orchestrator: it designs the change, dispatches specialist subagents to build and verify it, and owns the pull-request review cycle. The boundary between worker and orchestrator is the `git commit` — workers commit locally and hand back; the orchestrator pushes, opens the PR, and drives every reviewer finding to resolution.
28
29	The complexity of the work decides which mode you get. A one-line fix never pays the multi-agent tax; a multi-domain feature with shared state and lifecycle transitions gets the full ensemble.
30
31	## The ingredients
32
33	> Want the whole picture first? [`ARCHITECTURE.md`](ARCHITECTURE.md) lays out the system narrative and three Mermaid diagrams — the threading model, the agent ensemble, and the parallel pipeline. The stages below are the same machine, described in prose.
34
35	`/wizard` isn't just a prompt — it's a workflow built on specific ingredients that work together.
36
37	Carried over from v1 (still the foundation):
38
39	1. `CLAUDE.md` — your project's rules file. Coding standards, naming conventions, architecture decisions, anything Claude should always know. `/wizard` r

People who liked this also viewed

6 picks

Type

components-helperBased on the docs to provide code prompt files for vue component libraryPromptsNov 202312k36
VibeCode-Protocol-SuiteA comprehensive suite of protocols, meta-prompts, and orchestration tools designed to streamline software development workflows, project management, and team collaboration. Includes the VibeCode Protocol Suite with web interface, user manuals, and reusable templates for efficient development processes.PromptsJul 20262.3k24
Promptable-web-sdkWeb SDK for Promptable Website.PromptsDec 20221.3k37
LLMFlexA python package for developing AI applications with local LLMs.PromptsJan 2025151150
cli-repo-to-promptexport any codebase to a single LLM-ready markdown promptPromptsJul 202610653
Files2PromptVS Code extension to copy files for an LLM promptPromptsApr 20255923

claude-wizard

64stars20forks

View on GitHub

How to install claude-wizard?

vlad-ko/claude-wizard

$git clone https://github.com/vlad-ko/claude-wizard

Installs into the current project.

Summary

From one disciplined developer to an orchestrated team — without giving up the discipline.

Files · 1

View on GitHub

README.md

1	# Wizard v2: the multi-agent workflow
2
3	From one disciplined developer to an orchestrated team — without giving up the discipline.
4
5	Wizard v1 turned Claude Code from a fast coder into a careful one: it read before writing, tested before implementing, and attacked its own code before committing. That's still the foundation. v2 adds a second gear. When the work is big enough, the careful thread stops being the builder and becomes the orchestrator of a team of specialist agents — an issue maintainer who turns ideas into well-formed issues, an architect who designs, builders who implement in parallel, adversarial critics who try to break it, an independent reviewer that catches what the builders missed, and a review gate that won't let anything merge with an unanswered finding. One thread becomes a pipeline that drives a whole cohort of work to merge-ready at once.
6
7	If v1 was "think before you code," v2 is "design, then delegate, then verify — in parallel, without idling." And the most important shift is for you: you stop being a task-giver and become a conductor.
8
9	## Your role: conductor, not task-giver
10
11	This is the defining change of v2. You are not here to write detailed task breakdowns, spell out implementation steps, or write code. The agents do that. Your job is to keep the flow moving from idea → issue → PR → production — and nothing below that altitude.
12
13	Concretely, the human conductor:
14
15	- Sets direction. You bring the idea ("we need transfer-status tracking") — not the task list. Turning that idea into a structured issue with acceptance criteria is the issue-maintainer agent's first step, not your chore.
16	- Makes the product and judgment calls. When an agent hits an ambiguous requirement, a tradeoff, or a "which behavior is correct?" fork, that's yours to answer. The ensemble is deliberately good at building the thing right and deliberately defers deciding what's right to you.
17	- Unblocks. When the orchestrator pings that it's gated on a decision, a credential, or an external dependency, you clear it — then it resumes the whole cohort.
18	- Merges. The orchestrator drives every PR to merge-ready and declares it exactly once; you do the final merge. v2 never auto-merges. The merge button is the one piece of the cycle that stays human.
19
20	Everything between "idea" and "merge button" — issue authoring, design, building, reviewing, answering review findings — is delegated. You conduct; the ensemble plays.
21
22	## What it is
23
24	`/wizard` is a Claude Code [skill](https://docs.anthropic.com/en/docs/claude-code/skills) — a markdown playbook that changes how Claude operates for the duration of a task. v2 runs in two modes:
25
26	- Direct mode — a single thread runs the whole 8-phase lifecycle itself (this is v1's behavior, kept intact).
27	- Delegated mode — the thread you're talking to becomes an orchestrator: it designs the change, dispatches specialist subagents to build and verify it, and owns the pull-request review cycle. The boundary between worker and orchestrator is the `git commit` — workers commit locally and hand back; the orchestrator pushes, opens the PR, and drives every reviewer finding to resolution.
28
29	The complexity of the work decides which mode you get. A one-line fix never pays the multi-agent tax; a multi-domain feature with shared state and lifecycle transitions gets the full ensemble.
30
31	## The ingredients
32
33	> Want the whole picture first? [`ARCHITECTURE.md`](ARCHITECTURE.md) lays out the system narrative and three Mermaid diagrams — the threading model, the agent ensemble, and the parallel pipeline. The stages below are the same machine, described in prose.
34
35	`/wizard` isn't just a prompt — it's a workflow built on specific ingredients that work together.
36
37	Carried over from v1 (still the foundation):
38
39	1. `CLAUDE.md` — your project's rules file. Coding standards, naming conventions, architecture decisions, anything Claude should always know. `/wizard` r

People who liked this also viewed

6 picks

Type