| name | fable |
| description | Turn a concept, domain, or idea into a memorable fable (寓言) — a concrete allegorical story (situation → characters → conflict → resolution → moral) whose moral maps faithfully back to the concept. Use when the user wants to truly understand, teach, or explain something and would benefit from a story that sticks and spreads, rather than a flat explanation. Especially good for abstract, counterintuitive, or "hard to feel" concepts. Not for exact syntax or step-by-step procedures where precision matters. |
Fable 寓言
A fable is not a cute analogy. It is the oldest, most transmissible teaching form humans ever invented — and it has a structure. The job of this skill: take an abstract concept and cast it as a concrete allegory whose moral maps faithfully back to the concept.
When to use
- The user asks to understand / explain / teach a concept, and a story would help it stick.
- The concept is abstract, counterintuitive, or hard to build intuition for (e.g., eventual consistency, gradient descent, compound interest, entropy).
- The user wants something they can retell to others.
When NOT to use
- The user needs exact syntax, a precise spec, or a step-by-step procedure — give that directly. A fable that distorts precision is worse than no fable.
- The concept is too thin to carry narrative stakes (a single config flag, a trivial fact). Say so honestly instead of forcing one.
- The user is debugging or needs a factual fix — be direct, not literary.
How it gets invoked
By default reactive: the user asks to understand/explain a concept, or runs /fable <concept>. Proactively offer (one line, then wait — never dump a long fable unsolicited) when you detect an abstract/counterintuitive concept or repeated confusion（"还是没懂 / 为什么"）. Full protocol + output options (genre / length / audience / language): references/invocation.md.
Two modes — match the form to what you're teaching
- 寓意式 (moral fable) — for an abstract principle (复利, 熵, 局部最优): situation → conflict → resolution → one moral. The 7-step craft below produces this.
- 模型式 (structural allegory) — for a domain or mechanism (filesystem, TCP handshake, garbage collection): a complete, faithful, navigable model where every part maps to a real component, and the learner reasons about the model to derive behavior. Full craft in
references/fable-craft.md; worked example: the Great Archive (filesystem) in references/examples.md.
Rule of thumb: one idea to feel → 寓意式; a system to understand and predict → 模型式.
The craft (core recipe)
- Distill the principle. What is the one core idea the reader must walk away with? One moral per fable — never two.
- Find the allegory. A concrete situation whose internal logic mirrors the concept's logic. The mapping must be structural, not decorative.
- Cast characters. Concrete, vivid, ideally archetypal (the fox, the apprentice, the two kingdoms). Anthropomorphism is allowed only if it sharpens the idea — never if it misleads.
- Build the conflict. The tension in the story must be the concept's tension, dramatized — not a conflict pasted on top.
- Resolve. The ending must force the principle to surface naturally, without lecturing.
- State the moral, then map it back. End with an explicit 「寓意 ↔ 概念」mapping so the reader can verify it term by term.
- Fidelity test (mandatory, do not skip). Could this exact fable plausibly serve a different concept? If yes → it is too generic → redo. The allegory must be specific to THIS concept.
Language-quality loop (mandatory — this is the core)
A correct draft is not enough; the prose must be alive. Never ship a first draft. Full method in references/language-quality.md; in short:
- Calibrate — re-read one master in the target genre (Aesop / 庄子 / La Fontaine). Feel the temperature gap.
- Critique the draft against the quality checklist（terseness · concrete-not-abstract · no preaching · real stakes · one image at 120% · anti-slop）.
- Revise — cut 20%, replace every abstract label with a concrete image, kill every sentence that preaches.
- Re-run the fidelity test — revision must not break the 寓意↔概念 mapping.
Genre palette
Pick the tradition that fits the concept's texture:
- 伊索式 (Aesopic animal fable) — short, punchy, single sharp moral. Best for clean cause-and-effect principles.
- 庄子 / 韩非式 (Chinese 寓言) — philosophical, often paradoxical, sometimes unsettling. Best for counterintuitive or systems-level ideas.
- Parable — human-scale, ethical stakes. Best for ideas about behavior, incentives, or trade-offs.
Anti-slop (quality bar — non-negotiable)
- Restrained, vivid prose. No purple-gradient language, no emoji-as-icons, no forced cuteness, no SVG-drawn characters.
- One detail at 120%, everything else at 80%. A single image carries the weight; the rest stays quiet.
- If the concept resists a clean fable, say so — never distort the concept to make the story neat. Honesty over charm.
- The moral must be unambiguous. (Current AI struggles to extract morals from fables — so make ours extractable on purpose.)
Output shape (suggested)
【寓言】 <title>
<the story — concrete, short, with real stakes>
【寓意】 <one-line moral>
【寓意 ↔ 概念 映射】
- <story element> ↔ <concept element>
- ...
References
references/pedagogy.md — why fables teach (Egan, Bruner, empirical memory studies, Aesop as natural experiment).references/fable-craft.md — how to craft them: full manual, genre templates, fidelity test, anti-slop checklist, concept→genre routing.references/language-quality.md — the core: how to make prose actually good (8 principles, master study, contrastive calibration, draft→critique→revise loop, quality checklist).references/examples.md — 7 moral fables across three genres + 1 模型式 structural allegory (文件系统/大档案馆) + 1 honest-refusal case (幂等性), each with its fidelity check. The quality bar.references/invocation.md — when/how to fire: reactive default, proactive offer (never unsolicited dump), output options, /fable command.
Status: v0.2. Craft, language quality, invocation, and the distributable plugin layout are all in place. Next: end-to-end test in a live Claude Code session + README visual polish (see README roadmap).
