Dynamic Workflows logo

Dynamic Workflows

Organization
TheBeardedBearSAS
dynamic-workflows

Orchestrate dozens-to-hundreds of subagents from a script Claude writes (Claude Code Dynamic Workflows, trigger `ultracode`). Use when a task exceeds a single agent's context or needs more than ~4 concurrent workers — large audits, migrations, multi-source research, fan-out reviews. Distinct from Agent Teams (synchronous, ≤4 workers) and ralph-run (sequential single-context loop).

Overview

PublisherTheBeardedBearSAS
Repositoryclaude-craft
Skill namedynamic-workflows
Stars
105
Forks
9
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by TheBeardedBearSAS on GitHub. Read the source before you install it.

Installation

Install the Dynamic Workflows AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/TheBeardedBearSAS/claude-craft.git /tmp/claude-craft
mkdir -p .claude/skills
cp -r /tmp/claude-craft/Dev/i18n/base/Common/skills/dynamic-workflows .claude/skills/dynamic-workflows
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Dynamic Workflows in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Dynamic Workflows on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Dynamic Workflows is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Dynamic Workflows — orchestration multi-agents scriptée

Dynamic Workflows (Claude Code 2.1.154+, mot-clé déclencheur ultracode) laissent Claude écrire un script JavaScript qui orchestre des sous-agents de façon déterministe (boucles, conditions, fan-out). Le script tourne en arrière-plan ; on suit l'exécution via /workflows.

⚠️ Opt-in coûteux : un workflow peut lancer des dizaines à centaines de sous-agents. Ne l'utiliser que sur demande explicite (mot « ultracode », « use a workflow », « fan out agents ») ou quand le palier de complexité ci-dessous est franchi.

Quand l'utiliser — les 3 paliers d'orchestration

PalierOutilConcurrenceQuand
1Sub-agent simple (Task/Agent)1Une investigation isolée, garder le contexte principal propre
2Agent Teams (/team:audit, /team:sprint, /team:security)1 leader + ~3 workersTravail parallèle borné, synchrone, qui tient en un round
3Dynamic Workflows (ultracode)jusqu'à ~1000 sous-agents (cap concurrent ~16)La tâche dépasse le contexte d'un agent OU exige >4 workers OU une boucle/pipeline déterministe

Heuristique break-even : dès que tu écrirais « lance N agents, puis pour chacun fais X, puis agrège » — c'est un workflow. Si N ≤ 4 et une seule passe suffit, reste sur Agent Teams.

Patterns composables (prompts minimaux)

  • Fan-out & synthesize — N lecteurs en parallèle sur des sous-systèmes disjoints → un agent synthétise. (audit, cartographie de code)
  • Adversarial verification — chaque finding passé à ≥1 sceptique chargé de le réfuter ; ne garder que les CONFIRMED. Le pattern qualité le plus important pour la génération autonome.
  • Generate-and-filter — générer large (idées, candidats), puis filtrer par un juge.
  • Classify-and-act — un classifieur route chaque item vers le traitement adapté.
  • Pipeline — chaque item traverse toutes les étapes sans barrière (latence = pire chaîne, pas somme des étapes). C'est le défaut multi-étapes.
  • Loop-until-done / loop-until-dry — relancer des chercheurs jusqu'à K rounds sans nouveauté (découverte de taille inconnue : bugs, edge cases).

Squelette type (review → verify)

js
const results = await pipeline(
  DIMENSIONS,
  d => agent(d.prompt, { phase: 'Review', schema: FINDINGS }),
  review => parallel(review.findings.map(f => () =>
    agent(`Adversarially verify: ${f.title}. Default refuted=true if unsure.`,
          { phase: 'Verify', schema: VERDICT }).then(v => ({ ...f, verdict: v }))))
)
const confirmed = results.flat().filter(Boolean).filter(f => f.verdict?.isReal)

pipeline() par défaut (pas de barrière) ; parallel() seulement quand l'étape N a besoin de TOUS les résultats de N-1 (dédup, early-exit, comparaison croisée). schema force une sortie structurée validée. Pour des éditions concurrentes de fichiers, isolation: 'worktree'.

Monitoring

  • /workflows : progression live (phases, agents, tokens).
  • Le script revient avec un résultat agrégé ; lire ce résultat, décider la suite (souvent plusieurs workflows en séquence : comprendre → concevoir → implémenter → revoir).

À ne pas confondre

  • ralph-run (/common:ralph-run) : boucle séquentielle mono-contexte jusqu'à une DoD — pas de parallélisme. Pour une tâche qui dépasse ce modèle, basculer sur un Dynamic Workflow.
  • /effort ultracode : palier d'effort CLI (débit code max sur Opus 4.8) — orthogonal au déclencheur ultracode des workflows, même si souvent utilisés ensemble.

Référence : Claude Code Workflows · voir aussi @.claude/commands/common/sub-agents-patterns.md (tableau comparatif des orchestrations).

Frequently asked questions

What does the Dynamic Workflows AI skill do?

Orchestrate dozens-to-hundreds of subagents from a script Claude writes (Claude Code Dynamic Workflows, trigger `ultracode`). Use when a task exceeds a single agent's context or needs more than ~4 concurrent workers — large audits, migrations, multi-source research, fan-out reviews. Distinct from Agent Teams (synchronous, ≤4 workers) and ralph-run (sequential single-context loop).

Why use Dynamic Workflows on TypingMind?

Because you install it once and use it with any model. Dynamic Workflows is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Dynamic Workflows in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/TheBeardedBearSAS/claude-craft/tree/main/Dev/i18n/base/Common/skills/dynamic-workflows. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Dynamic Workflows?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Dynamic Workflows?

As many as you like. As long as a model supports skills, you can use Dynamic Workflows with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Dynamic Workflows AI skill free?

Yes. It is published on GitHub by TheBeardedBearSAS under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇