Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add hanzhcn/laohan-skills --skill laohan-yinsegit clone --depth 1 https://github.com/hanzhcn/laohan-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/hanzhcn/laohan-skills/laohan-yinse)<a href="https://agentmods.dev/skills/hanzhcn/laohan-skills/laohan-yinse"><img src="https://agentmods.dev/badge/skills/hanzhcn/laohan-skills/laohan-yinse/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/hanzhcn/laohan-skills/laohan-yinse"><img src="https://agentmods.dev/badge/skills/hanzhcn/laohan-skills/laohan-yinse.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00110 | $0.09246 |
| Opus 5.5 | $0.00044 | $0.03698 |
| Sonnet 5.5 | $0.00022 | $0.01849 |
| Haiku 4.5 | $0.00011 | $0.00925 |
Grade A, and why
laohan-yinse scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 256 lines — stays where its author put it; the contents beside it link to each section on GitHub.
laohan-yinse — 复刻博主声音模型（服务器微调）
使命与合同
每个复刻博主一个专属声音模型，唯一方法 = pro6000 算力机 CosyVoice3 微调训练（Jeffrey 2026-10-27 钉死：不走零样本）。声音模型是博主复刻的核心资产：数字人=形象克隆+口型驱动，声音来自 TTS——没有本人音色模型，出来的是"通用音色配这张脸"，瞬间违和。
硬性前置是本博主已有训练、登记并验收的专属模型，不是每期重新训练。同一博主/形象身份固定默认模型版本；换集数、稿件、服装、场景或参考情绪不换模型、不重训。不能跨博主混用，也不能把模型的历史接受当本期新WAV已通过整轨听审。
方法依据：~/Documents/声音资产/Jeffrey/训练方法论.md §3/§4/§6—8；整稿声音与口型衔接读 数字人专属音色融合方法.md 与 laohanvideo/motion/mastery/DH-ROUTE.md。复用数据、训练与听审方法，不复用Jeffrey个人权重、参考池或接受记录；本博主必须有自己的资产绑定。
这些是Jeffrey/Motion项目的方法参考；其v3-05、历史首单、主干接入/发布及提前候选授权不转移到本项目。当前创作者、验收人和制作状态按本项目合同/STATUS；Jeffrey是负责人及听看验收人，不是被选博主。
素材标准
| 项 | 要求 |
|---|---|
| 总量 | 目标8–10分钟净有效人声（按清洗后逐段时长统计，不以原视频容器时长代替） |
| 质量 | 单人清晰口播、领夹麦/近场；无明显 BGM/混响（BGM 混入训练=音色污染） |
| 格式 | 训练用24kHz单声道PCM16 WAV；原视频另存，不反复重采样 |
| 同源 | 与形象素材同一博主 |
数据铁律（源自训练方法论 v3.1 终版，codex 三轮审查——2026-10-28 三轮查漏补录）
- 逐段音文校对：默认人工，覆盖全部保留段；用户明确委派时按下方角色与委派分支如实执行，不能代勾人工页。ASR原始输出、单纯文本多数票或schema合法不能直接进训练；抽校不够（Jeffrey判例：错词污染音色）。
- 防泄漏分组切分：先按录音来源（视频条）去重分组，同源段只进同一侧 split（相邻/重叠内容跨 train/dev=cv 虚假乐观）——禁止随机 shuffle 9:1
- 采样率统一 24000Hz mono（conf 固定参数；上传前
ffmpeg -ar 24000 -ac 1） - 读法一致化：数字/缩写/多音字写实际读法（英文：Dr./don't/年份等统一）
- 宁缺毋滥：口误、重复、语气词泛滥段剪掉
- 训练期生产监控：与生产 TTS 桥共卡，训练期监控生产延迟/错误率，超标暂停训练
- 产物自检：特征三件套（utt2embedding.pt/spk2embedding.pt/utt2speech_token.pt）齐全、parquet ID 集合==wav.scp、记录实际进训练样本数与名义对账
- num_workers=1（小数据防 shard 复制双倍曝光）；checkpoint 合并用官方 average_model（val_best 不稳降 --num）
验收（v3.1 口径修正，2026-10-28）
验收 = A/B 对照 + Jeffrey 盲听：原版（零样本）与微调版在相同参考/文本/固定采样下对照，查错漏字/重复/自然度/相似度，Jeffrey 盲听终审拍板。F0 是辅助分布指标，不做窄区间硬门槛（此前"F0 ±20Hz"口径作废——与真源冲突）。
记录并固定本次可控采样参数和随机种子；匿名化试听文件名、随机播放顺序并另存揭盲映射，听审后再揭盲。固定种子不保证跨版本/硬件逐样本确定；连续稳定性检查须有独立生成样本，不能反复播放同一WAV冒充。
验收角色与委派
默认由Jeffrey听审。仅当用户实际指令明确委派了具名阶段，才可由受托方在原范围内裁决，并记录原指令、创作者/本期输入SHA、阶段、实际审音方式、结论及限制；不得写成Jeffrey亲听。源素材英文校对、模型选择、整轨听感、成片接受和发布分别核范围。一次性阶段委派不自动扩展；用户明确作出覆盖本系列后续新配音的持续委派时（概括性转述，实际范围以留存用户原文为准），按实际系列/阶段持续有效，不再逐轨反问或让不懂英语的用户承担默认听审。项目当前授权优先于本节通用默认，负责人自行安排检查、裁决与修正，逐条绑定新WAV证据；这不自动授权成片接受或发布。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago Changed · +10 lines 77761e83b71a
- 10d ago Changed · +108 lines · -50 tokens per session 5ca09b8b24a2
- 11d ago First seen · 138 lines · 160 tokens per session scan A 1eaf645eb7aa
laohan-yinse is a skill published in the GitHub repository hanzhcn/laohan-skills (15 stars, last pushed today), licensed MIT. It adds 110 tokens to every session and 9,246 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-28.
laohan-yinse compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| laohan-yinsehanzhcn/laohan-skills | Skill | 15 | today | MIT | A | 110 | Claude Code, Codex |
| baijiahao-uploaddreammis/social-auto-upload | Skill | 16k | 16d ago | MIT | A | 49 | Claude Code, Codex |
| hupu-uploaddreammis/social-auto-upload | Skill | 16k | 16d ago | MIT | A | 42 | Claude Code, Codex |
| douyin-uploaddreammis/social-auto-upload | Skill | 16k | 16d ago | MIT | A | 93 | Claude Code, Codex |
| xiaohongshu-uploaddreammis/social-auto-upload | Skill | 16k | 16d ago | MIT | A | 95 | Claude Code, Codex |
| Freelance EmpireLeoYeAI/openclaw-master-skills | Skill | 2.2k | 2d ago | MIT | A | 71 | Claude Code, Codex |
| ad-readyLeoYeAI/openclaw-master-skills | Skill | 2.2k | 2d ago | MIT | A | 191 | Claude Code, Codex |