| name | mythos-scaffold |
| description | Activate Mythos scaffold. Fable-distilled working patterns (decomposition, verification, next-action decision) plus a discipline framework (mode, tool-stack, context-priming, decomposition, agent-loop, verification, failure-recovery) and domain modes (research, migration) that push Opus-class and smaller models toward Mythos/Fable-grade agentic behavior. Use when starting complex multi-step work, long-horizon research, codebase migration, or tasks that benefit from systematic persistence and verification. Do NOT fully load on Fable 5 / Mythos 5; over-scaffolding degrades Mythos-class models. |
Mythos Scaffold
Pattern-based agentic discipline framework. Brings Plan-and-Execute, Reflexion, Self-Refine, and Ralph Loop into structured protocols, anchored by working patterns distilled from Claude Fable 5 itself on 2026-07-06, during the 2-7 July free window.
Honest expectation: the scaffold transfers process discipline, not capability. It makes Opus 4.8 a disciplined Opus 4.8. Raw reasoning depth, long-horizon coherence, and novel pattern recognition live in the weights and cannot be patched in via prompts.
Model gating (check first)
| Session model | What to load |
|---|
| Opus 4.8 / Opus 4.x / Sonnet / Haiku | Tiered activation below |
| Fable 5 / Mythos 5 | Nothing, or at most core/fable-distilled.md section 2 (verification). State goal and constraints; do not enumerate steps. Per Anthropic's migration guide, prior-model scaffolding reduces Fable-class output quality. |
Tiered activation
Always load (cheap, covers most sessions):
- Distilled patterns:
core/fable-distilled.md
- Mode rules:
core/mode.md
Load on trigger only, not preemptively:
Domain Modes
For specialized tasks, also load the relevant domain bundle:
- Long-horizon research synthesis:
domains/research/
mode.md -> retrieval.md (6-tier source hierarchy) -> synthesis.md (claim graph, contradiction tree) -> cite-verify.md (hallucination check) -> output.md (academic, blog, brief, slide, wiki)
- Codebase migration / framework upgrade:
domains/migration/
mode.md -> audit.md (footprint, breaking change matrix) -> plan.md (phasing, calibrated time estimates) -> execute.md (atomic commits, sub-agent parallel) -> rollback.md (4 strategies + drill protocol)
Slash Command
The /mythos-mode <task> command (installed at ~/.claude/commands/mythos-mode.md) runs a full Mythos pass on a single task. Use it when the task is well-defined and you want the scaffold to drive end-to-end.
When to Use
- Multi-step tasks where the model might stall, drift, or miss verification
- Long-horizon research across many sources (research domain)
- Framework or codebase migrations needing audit plus phased rollout (migration domain)
- Debugging cycles where naive retry fails (failure-recovery / Ralph Loop)
- Tasks where sub-agent parallelism helps (decomposition, with explicit model routing)
When NOT to Use
- Single tool call, single answer questions: overhead not worth it
- Tasks already covered by a more specific skill
- Sessions running on Fable 5 / Mythos 5 (see model gating above)
Changelog
- 2026-07-06:
core/fable-distilled.md added (patterns written by Fable 5 about its own decomposition, verification, and next-action habits). Read-all-8 activation replaced with tiered loading. Model gating added. Stale Opus 4.7 / hardcoded date references fixed. Sub-agent model routing added to decomposition.
- 2026-04-27: initial 6-skill framework plus research and migration domains.
Source
https://github.com/kasparovabi/claude-mythos-scaffold (MIT, by kasparovabi). Inspired by Claude Mythos Preview behaviors, not affiliated with Anthropic.
