Fable Mode — Fable 5.1 (v3, agent-routed)
Same structural delegation as fable-opus: the orchestrator is a real agent
definition (agents/fable-orchestrator.md) with no Write/Edit tool, so it
cannot do the work inline. The only difference from fable-opus is the model the
orchestrator runs on. Fable 5.1 is the top of the ladder (Haiku → Sonnet → Opus
→ Fable → user); gaps it cannot close go to the user.
Fable 5.1's characteristic gaps under this discipline: none measured
(benchmarked 2026-09-02, 16 paired runs, n=2 per cell: 100/100 in every cell,
both arms; skill cost ~1.1× tokens, ~2× wall). This is a provenance runner —
worklog, reproducible checks, flagged judgment calls — not an accuracy runner.
The orchestrator definition keeps the frontier-tier guards (external-artifact
verification, scope rule, replan budget) because those produce the trail.
If a task has one obvious correct approach and fits in a single pass, skip this
loop and do it directly. On a Fable-class model the stage map is ceremony for
anything short of multi-session or many-file work.
How to run it
- Confirm the fable agents are installed (
fable-orchestrator,
fable-worker-sonnet, fable-worker-haiku, fable-verifier appear in the
available agent types) AND the Agent/Task tool accepts model: "fable".
- Agents missing → fall back to the inline method: spawn a general-purpose
agent with
model: "fable" and pass it the Core loop and operational
rules verbatim from agents/fable-orchestrator.md.
"fable" not accepted as a model → say so and run fable-opus instead.- Per-invocation
model override is documented for Claude Code
(code.claude.com/docs/en/sub-agents: "the per-invocation model
parameter" outranks the definition's frontmatter; fable is a listed
alias) and confirmed on the Cowork Agent tool. On a runtime that lacks
it, copy agents/fable-orchestrator.md to
agents/fable-orchestrator-fable.md, set name: fable-orchestrator-fable
and model: fable, and route to that instead.
- Spawn @fable-orchestrator via the Task tool (
subagent_type: "fable-orchestrator", model: "fable"). Brief it with: the user's task,
the output directory, relevant session context, and any user-set limits
(warning threshold, worker cap, deadline).
- Do not restate the Core Loop or operational rules in the briefing — the
orchestrator's agent definition carries them. Brief the task, not the
method.
- When it returns, relay the result, every stage it marked unverified, and its
recommendations (surfaced scope it did not build).
- Mandatory delivery gate: before presenting the result to the user, invoke
the double-check skill on the finished deliverable. If the orchestrator's
fable-verifier already cold-checked the final document (not just
intermediate outputs), double-check will detect that and run only the
synthesis-seam check instead of a full panel — its own rules handle this. Do
not skip the gate; the user relies on it instead of re-checking the work
themselves.
Known limitation
Same side door as fable-opus: the orchestrator keeps Bash for verification
commands, and Bash can technically create files. Its definition forbids that
use.
1---2name: fable-fable3description: Run fable-mode execution discipline on Claude Fable 5.1 — the top of the escalation ladder and the strongest staged run available. Routes the task to the @fable-orchestrator agent with its model overridden to Fable (`model: "fable"`), which stages the work, delegates artifact production to @fable-worker-sonnet / @fable-worker-haiku, and cold-checks deliverables with @fable-verifier. Trigger when the user explicitly asks for thorough/systematic/"deep work" handling on the strongest model ("fable on fable", "stage this on fable", "deep work mode, fable", "run this on 5.1"), or when a fable-opus run reports the task exceeds Opus's ceiling. Benchmarked 2026-09-02: zero correctness delta on Fable 5.1 at ~1.1× tokens, ~2× wall — a provenance runner, not an accuracy runner. Do NOT use for ordinary single-pass tasks — prefer fable-opus, fable-sonnet, or fable-haiku when the task doesn't need peak reasoning. Fable is plan-included; bulk work still routes down-tier.4---56# Fable Mode — Fable 5.1 (v3, agent-routed)78Same structural delegation as fable-opus: the orchestrator is a real agent9definition (`agents/fable-orchestrator.md`) with no Write/Edit tool, so it10cannot do the work inline. The only difference from fable-opus is the model the11orchestrator runs on. Fable 5.1 is the top of the ladder (Haiku → Sonnet → Opus12→ Fable → user); gaps it cannot close go to the user.1314Fable 5.1's characteristic gaps under this discipline: none measured15(benchmarked 2026-09-02, 16 paired runs, n=2 per cell: 100/100 in every cell,16both arms; skill cost ~1.1× tokens, ~2× wall). This is a provenance runner —17worklog, reproducible checks, flagged judgment calls — not an accuracy runner.18The orchestrator definition keeps the frontier-tier guards (external-artifact19verification, scope rule, replan budget) because those produce the trail.2021If a task has one obvious correct approach and fits in a single pass, skip this22loop and do it directly. On a Fable-class model the stage map is ceremony for23anything short of multi-session or many-file work.2425## How to run it26271. Confirm the fable agents are installed (`fable-orchestrator`,28 `fable-worker-sonnet`, `fable-worker-haiku`, `fable-verifier` appear in the29 available agent types) AND the Agent/Task tool accepts `model: "fable"`.30 - Agents missing → fall back to the inline method: spawn a general-purpose31 agent with `model: "fable"` and pass it the Core loop and operational32 rules verbatim from `agents/fable-orchestrator.md`.33 - `"fable"` not accepted as a model → say so and run fable-opus instead.34 - Per-invocation `model` override is documented for Claude Code35 (code.claude.com/docs/en/sub-agents: "the per-invocation `model`36 parameter" outranks the definition's frontmatter; `fable` is a listed37 alias) and confirmed on the Cowork Agent tool. On a runtime that lacks38 it, copy `agents/fable-orchestrator.md` to39 `agents/fable-orchestrator-fable.md`, set `name: fable-orchestrator-fable`40 and `model: fable`, and route to that instead.412. Spawn **@fable-orchestrator** via the Task tool (`subagent_type:42 "fable-orchestrator"`, `model: "fable"`). Brief it with: the user's task,43 the output directory, relevant session context, and any user-set limits44 (warning threshold, worker cap, deadline).453. Do not restate the Core Loop or operational rules in the briefing — the46 orchestrator's agent definition carries them. Brief the task, not the47 method.484. When it returns, relay the result, every stage it marked unverified, and its49 recommendations (surfaced scope it did not build).505. **Mandatory delivery gate:** before presenting the result to the user, invoke51 the **double-check** skill on the finished deliverable. If the orchestrator's52 fable-verifier already cold-checked the *final* document (not just53 intermediate outputs), double-check will detect that and run only the54 synthesis-seam check instead of a full panel — its own rules handle this. Do55 not skip the gate; the user relies on it instead of re-checking the work56 themselves.5758## Known limitation5960Same side door as fable-opus: the orchestrator keeps Bash for verification61commands, and Bash can technically create files. Its definition forbids that62use.
mrtooher