fable-orchestrationlisted

# Fable 5.1 Orchestration Pay Fable rate only for judgment. Everything else runs on cheaper lanes. **Rule:** Fable decides. Opus 5 builds. Sonnet 5 grinds. Haiku 4.5 scouts. Fresh-context verifier checks. Fable earns its rate on ambiguous scope and parallel builds, not on fully specified single-builder tasks. Facts + sources: `reference/models.md`. Configs: `reference/wiring.md`. Paste-in prompts: `reference/prompt-kit.md`. Symptom -> fix: `reference/failure-modes.md`. Drop-in `settings.json` and `CLAUDE.md` examples: `${CLAUDE_PLUGIN_ROOT}/settings/`. ## Lanes Agents ship with this plugin as `fable-orchestration:<name>`. | Lane | Agent | Model | Effort | Job | |---|---|---|---|---| | Architect | main session | `fable` ($10 / $50, cache read $0.25) | `high` | Plan, ambiguous calls, root cause, final judgment | | Architect, one-off | `fable-architect` | `fable` | `high` | Plan-only pass from an Opus/Sonnet main. Returns text, no code | | Builder | `opus-builder` | `opus` ($5 / $25) | `high` | Multi-file implementation, hard debugging | | Worker | `sonnet-worker` | `sonnet` ($2 / $10) | `medium` | Boilerplate, tests, migrations, docs, research reads | | Scout | `haiku-scout` | `haiku` ($1 / $5) | none (no effort levels) | Locate files, grep, condense logs | | Verifier | `fresh-verifier` | `opus`, clean context | `high` | Check work vs spec. Never the builder | Anthropic's Fable 5.1 guidance: `low` effort on Fable often exceeds prior models at `xhigh`. Cheaper lanes are f