
fable-orchestration-5.1
Claude Code skill: Fable 5.1 decides, Opus 5 / Sonnet 5 / Haiku 4.5 execute. Five wiring patterns (advisor, architect+delegate, agent teams, workflows, /goal…
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up fable-orchestration-5.1 (claude-plugin project) into my current project. Found on https://claudeers.com/fable-orchestration-51 Repo: https://github.com/y0f/fable-orchestration-5.1 Homepage/docs: — Detected install method: claude-plugin → /plugin install fable-orchestration-5-1@y0f/fable-orchestration-5.1 Category: skills. Platforms: cli. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: active; community-verified: false. Confirm the source before running anything.
/plugin marketplace add y0f/fable-orchestration-5.1 /plugin install fable-orchestration-5-1@y0f/fable-orchestration-5.1
git clone https://github.com/y0f/fable-orchestration-5.1
// compatibility
| Platforms | cli |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | MIT |
| Pricing | open-source |
| Language | Shell |
fable-orchestration
Fable 5.1 decides. Opus 5 builds. Sonnet 5 grinds. Haiku scouts.
Claude Code plugin. Pay Fable rate for judgment only.
flowchart LR
U([task]) --> F
subgraph FABLE["Fable 5.1 · $10 / $50"]
F[architect<br/>plan · decide · judge]
end
F -. spawn async .-> B & W & S
B & W & S -- result --> F
F -. verify .-> V
V -- PASS / FAIL + evidence --> F
F --> D([done])
subgraph LANES["cheap lanes"]
B[opus-builder<br/>$5 / $25]
W[sonnet-worker<br/>$2 / $10]
S[haiku-scout<br/>$1 / $5]
V[fresh-verifier<br/>Opus, clean context]
end
classDef fable fill:#8A2BE2,color:#fff,stroke:none
classDef lane fill:#2b2d42,color:#fff,stroke:none
classDef io fill:#edf2f4,color:#2b2d42,stroke:#8d99ae
class F fable
class B,W,S,V lane
class U,D io
Install
Inside Claude Code:
/plugin marketplace add y0f/fable-orchestration-5.1
/plugin install fable-orchestration@fable-orchestration
Then merge settings/settings.example.json (this repo, or the installed copy under ~/.claude/plugins/marketplaces/fable-orchestration/settings/) into ~/.claude/settings.json. Run /model fable per session when the task warrants it. Agents appear as fable-orchestration:opus-builder etc.
Pick wiring
flowchart TD
Q1{unattended,<br/>checkable end state?} -- yes --> G["/goal <check> on top of the pattern below"]
Q1 -- no --> Q2
G --> Q2{same step over<br/>many files?}
Q2 -- yes --> WF["workflow<br/><i>use a workflow to …</i>"]
Q2 -- no --> Q3{peers must argue<br/>or own layers?}
Q3 -- yes --> AT["agent teams<br/>Fable lead · Sonnet peers"]
Q3 -- no --> Q4{few decisions,<br/>short session, /clear often?}
Q4 -- yes --> AD["advisor mode<br/>/model sonnet · /advisor fable"]
Q4 -- no --> AR["architect + delegate<br/>/model fable · lanes build"]
classDef q fill:#edf2f4,color:#2b2d42,stroke:#8d99ae
classDef a fill:#8A2BE2,color:#fff,stroke:none
class Q1,Q2,Q3,Q4 q
class G,WF,AT,AD,AR a
Advisor re-reads the whole transcript uncached at $10/MTok per consult. Architect reads its own context from cache at $0.25. Cost table in wiring.md.
Layout
| Path | What |
|---|---|
skills/fable-orchestration/SKILL.md | Loads on trigger. Lanes, wiring, effort, cost levers, traps, checklist. |
reference/models.md | Facts with sources. Pricing, advisor pairings, 5.1 vs 5 deltas. |
reference/wiring.md | Configs for all five patterns. Frontmatter, env, limits, cost table. |
reference/prompt-kit.md | 16 paste-in blocks for 5.1, each marked whether Claude Code's Fable system prompt already carries it (checked live). Old-prompt patterns to delete. |
reference/failure-modes.md | Symptom → cause → fix. |
agents/ | Five lanes, pinned model + effort + tools. |
settings/ | settings.json examples, CLAUDE.md routing block. |
bench/ | run.sh: same task on Fable solo, Opus solo, and orchestrated, 3 runs each, median reported. task-parallel.md is the shape orchestration claims to win; task-small.md the shape it should lose. |
Four traps
- Cyber-flagged prompt → session falls back to Opus 4.8 (one transcript notice, then stays there). Bio → Opus 5.
claude --safe-modefinds the trigger. - Unpinned subagents inherit Fable.
CLAUDE_CODE_SUBAGENT_MODEL=opus. - Subagent cache is 5 min.
subagentPromptCacheTtl: "1h". - A fork subagent inherits Fable and the whole transcript, ignoring
CLAUDE_CODE_SUBAGENT_MODEL. Spawn a named lane instead.
MIT. Each claim is attributed in models.md: Claude Code docs, the Anthropic announcement, or Anthropic's bundled Fable 5.1 migration guidance.
// faq
What is fable-orchestration-5.1?
Claude Code skill: Fable 5.1 decides, Opus 5 / Sonnet 5 / Haiku 4.5 execute. Five wiring patterns (advisor, architect+delegate, agent teams, workflows, /goal), effort routing, fallback traps, 16-block prompt kit, pinned agent lanes.. It is open-source on GitHub.
Is fable-orchestration-5.1 free to use?
fable-orchestration-5.1 is open-source under the MIT license, so it is free to use.
What category does fable-orchestration-5.1 belong to?
fable-orchestration-5.1 is listed under skills in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/fable-orchestration-51)
// retro hit counter
[](https://claudeers.com/fable-orchestration-51)
// reviews
// guestbook
// related in Claude Skills
An agentic skills framework & software development methodology that works.
💫 Toolkit to help you get started with Spec-Driven Development
AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, and more). Turn any folder of code, SQL schemas, R scripts, shell scripts, docs,…