Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Distills AI agent voice patterns from conversation into a Persona Voice Card v1.0 (self/user/language/register/emoji_policy/flourish only — no judgment-shaped fields), validates schema + content safety + differentiation, and routes contributions through the existing submission portal. Use when the user wants to crystallize a recurring voice/tone, fork an existing persona, or share one with the community.
.claude/skills/telagod-cultivating-personas/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 8% | 0% |
<!-- safety-scan: ignore TOOLS_PRIVILEGED 特权理由：Bash 用于 schema 校验脚本；Write/Edit 用于在用户确认后落盘 <slug>.json -->
> 人格是声音，不是判断。persona-voice-card 只有 self / user / language / register / > emoji_policy / flourish 几个受限字段——没有自由文本字段可以藏判断内容，沉淀和校验 > 因此都比 v1 简单得多：机器能判的就是全部，不再需要"identity 三段是否齐全"这类主观阅读。
| 模式 | 触发 | 入口 | |------|------|------| | distill | 会话中识别"魔尊形成稳定声音模式" | 主动提议→显式确认 | | create | 用户显式想造新人格 | 引导式问答→生成 card | | validate | 已有 card，校验 schema + 内容安全 + 差异度 | 一键扫描 |
| 场景 | 使用 | 理由 | |------|------|------| | 自称 / 称呼用户已稳定 ≥ 3 会话 | ✅ distill | 声音模式可结晶 | | 用户想 fork 内置人格做变体 | ✅ create | 直接生成骨架 | | 已有 card 但质量存疑 | ✅ validate | 客观闸门 |
skills/_kernel/ 或对应领域 skill 的事，不是人格的事（见下方"安全脊柱"）人格虽是文本，仍有红线：
additionalProperties:false，没有自由文本字段能装下判断内容，这是设计上的，不是靠自觉详细审查清单见 references/persona-safety.md。
| 你想做的 | 走哪卷 | |---------|--------| | 从会话沉淀人格 | references/distillation.md | | 校验 voice 差异度 | references/voice-consistency.md | | 通过 submit portal 提交 | references/publishing.md |
bash# 校验现有 card node scripts/persona_forge.js validate config/personas/<slug>.json # 从会话提炼（交互式，不直接落盘，先输出预览） node scripts/persona_forge.js distill --voice-hint "<观察笔记>" # 生成提交 payload（给 submit.html 用） node scripts/persona_forge.js publish config/personas/<slug>.json
<slug>.json 给魔尊 review参见姊妹 skill cultivating-skills——专司工程方法沉淀。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,651 | 9,294 | +64% | 1 | 1 | 0% | 809 | 1,330 | +64% | 0 | 0 | — |
case-02 | fail→fail | 17,548 | 9,437 | -46% | 1 | 1 | 0% | 2,424 | 2,247 | -7% | 0 | 0 | — |
case-03 | fail→pass | 16,417 | 10,451 | -36% | 1 | 1 | 0% | 2,598 | 2,738 | +5% | 0 | 0 | — |
case-04 | pass→pass | 18,940 | 13,924 | -26% | 1 | 1 | 0% | 2,720 | 3,083 | +13% | 0 | 0 | — |
case-05 | pass→pass | 17,440 | 12,665 | -27% | 1 | 1 | 0% | 3,412 | 3,747 | +10% | 0 | 0 | — |
case-06 | pass→fail | 19,812 | 16,158 | -18% | 1 | 1 | 0% | 3,084 | 3,482 | +13% | 0 | 0 | — |
case-07 | pass→pass | 16,662 | 9,121 | -45% | 1 | 1 | 0% | 2,691 | 2,401 | -11% | 0 | 0 | — |
case-08 | pass→pass | 15,466 | 5,780 | -63% | 1 | 1 | 0% | 2,015 | 1,798 | -11% | 0 | 0 | — |
case-09 | pass→pass | 17,440 | 8,550 | -51% | 1 | 1 | 0% | 2,478 | 2,164 | -13% | 0 | 0 | — |
case-10 | pass→pass | 9,267 | 4,333 | -53% | 1 | 1 | 0% | 1,493 | 1,655 | +11% | 0 | 0 | — |
case-11 | fail→pass | 12,927 | 7,783 | -40% | 1 | 1 | 0% | 1,785 | 2,108 | +18% | 0 | 0 | — |
case-12 | fail→pass | 14,419 | 4,088 | -72% | 1 | 1 | 0% | 2,207 | 1,561 | -29% | 0 | 0 | — |
case-13 | fail→pass | 14,531 | 10,871 | -25% | 1 | 1 | 0% | 1,769 | 2,482 | +40% | 0 | 0 | — |
case-14 | pass→pass | 18,555 | 10,806 | -42% | 1 | 1 | 0% | 2,554 | 2,549 | -0% | 0 | 0 | — |
case-15 | pass→pass | 12,968 | 4,778 | -63% | 1 | 1 | 0% | 1,923 | 1,662 | -14% | 0 | 0 | — |
case-16 | fail→pass | 17,805 | 11,963 | -33% | 1 | 1 | 0% | 2,442 | 2,635 | +8% | 0 | 0 | — |
case-17 | fail→pass | 18,474 | 4,783 | -74% | 1 | 1 | 0% | 2,653 | 1,751 | -34% | 0 | 0 | — |
case-18 | fail→pass | 19,005 | 6,133 | -68% | 1 | 1 | 0% | 2,562 | 1,912 | -25% | 0 | 0 | — |
case-19 | fail→pass | 18,670 | 4,884 | -74% | 1 | 1 | 0% | 2,286 | 1,715 | -25% | 0 | 0 | — |
case-20 | fail→pass | 10,210 | 2,641 | -74% | 1 | 1 | 0% | 1,446 | 1,365 | -6% | 0 | 0 | — |
case-21 | pass→pass | 16,699 | 12,381 | -26% | 1 | 1 | 0% | 2,365 | 2,853 | +21% | 0 | 0 | — |
case-22 | fail→pass | 15,132 | 11,727 | -23% | 1 | 1 | 0% | 2,162 | 2,934 | +36% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +41 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.