laohan-skills: Instructions file for Claude Code

CLAUDE.md

laohan-skills CLAUDE.md is an instructions file for Claude Code from hanzhcn/laohan-skills. It costs 3,470 tokens per session, scanned A, original, MIT.

A project instruction file for a Chinese-language talking-head video production workflow branded as laohanAI. It defines naming rules, the stages for planning and producing episodes, required project documents, and checks before an episode is accepted.

In plain words
What is it for?
Use it to guide episode planning, footage and source tracking, animation implementation, rendering, status updates, and final project validation.
Why use it?
It keeps a multi-step video workflow consistent and prevents outdated names, missing files, or skipped production checks.

Instructions file for Claude Code

Written for Claude Code: the file is CLAUDE.md. Also seen: mentions CLAUDE.md; mentions Codex.

This is hanzhcn/laohan-skills's own configuration. It tells Claude Code how to work on laohan-skills itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything laohan-skills configures →

Not installable: its command points at a path on the author’s own machine, so it runs nowhere else. The line is /Users/hanzhmacbookair/Documents/laohanAI视频创作/docs/动画效果方法论.md.

Reuse

Borrowing it

Nothing to install: this file belongs to hanzhcn/laohan-skills. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/hanzhcn/laohan-skills/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/hanzhcn/laohan-skills

Made for: Claude Code.

Wrote 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.

agentmods badge for laohan-skills CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/hanzhcn/laohan-skills/claude-md/github.svg)](https://agentmods.dev/instructions/hanzhcn/laohan-skills/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/hanzhcn/laohan-skills/claude-md"><img src="https://agentmods.dev/badge/instructions/hanzhcn/laohan-skills/claude-md/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.

agentmods 80×15 button for laohan-skills CLAUDE.md

Your own site · 80×15
<a href="https://agentmods.dev/instructions/hanzhcn/laohan-skills/claude-md"><img src="https://agentmods.dev/badge/instructions/hanzhcn/laohan-skills/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 3,470 This file is loaded in full into every session.
When invoked 3,470 The same file — it is already loaded in full.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.03470 $0.03470
Opus 5 $0.01735 $0.01735
Sonnet 5 $0.00694 $0.00694
Haiku 4.5 $0.00347 $0.00347

Measured 9d ago against content hash 2209f559fa36, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

laohan-skills CLAUDE.md 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.

CLAUDE.md · 48 lines

How it starts

The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.

laohan-skills project rules

Brand name

  • 对外品牌名统一为 laohanAI。所有口播稿、封面提示词、系列标签、安装器标题和后续生成内容不得再使用旧名“寒武纪AI”或“寒武纪说AI”。

Talking-head video workflow

真人口播动画默认采用 CODEX_DIRECT + Remotion:

  • 新episode的导演预制默认调用laohan-daoyan v2.1.2:它只读取视频项目当前method-v5.jsondirector-state-v5.md。第一请求先从内容发散2—3个在视觉隐喻、空间组织、运动逻辑或素材角色上本质不同的完整方向,方向形成后才按需读取最小范围Plugin知识,再统一收敛;第二请求只做最终复审,对真正值得增强的位置内部发散2—3个本质不同的候选。非穷尽技术搜索地图只提醒搜索维度,不是必选清单,未入选候选不落盘。两请求都停在WAITING_FOR_FOOTAGE。完成时必须同步本期_status.md,并在最后一次写入后实时运行bash scripts/check-episode-contract.sh episodes/<slug> config;不能用开始时PASS或干净Git代替episode准入。它不得剪辑、生成clean/SRT、写旧beat-sheet/EDL/motion-plan、实现Remotion或渲染。旧METHOD_LAB只在Jeffrey明确要求历史路线时读取laohan-daoyan/references/method-lab-legacy.md
  • ⑧—⑪由同一个 Codex 任务连续完成:完整观看原片并剪辑、写direct brief与按需source manifest、直接实现Remotion、渲染完整candidate并交Jeffrey观看。laohan-donghua、AST、双renderer与proposal/extension/seal只用于Jeffrey明确发起的METHOD_LAB或历史episode。
  • laohan-sucai/SKILL.md:⑩仅在 source manifest 有真实 PROOF/BROLL 请求时供应 provider/license/source/SHA 完整且 visually_verified 的本期素材;无请求标 not_applicable,不创建空素材任务。
  • laohan-bianpai/SKILL.md:全流程的唯一状态路由器;从 laohanAI视频创作 根目录运行,按 artifact + mechanical gate 报告唯一下一步。⑫—⑭在 laohan-yunying 定义前必须 BLOCKED。
  • V5旧期已有pending candidate再接入V5.1时,只能使用workflow受控迁移冻结原candidate与director-state SHA;bianpai视该期D1/D2为历史兼容完成并停在JEFFREY_REVIEW,不得倒退覆盖已生成样片。
  • episode 存在且通过 workflow verify-episode-supersession.mjssupersession-record.json 时,bianpai 必须优先返回 SUPERSEDED 并拒绝 vendors/status/next/check;不得覆盖旧 executor lock 或把作废期迁移到新 runtime。
  • laohan-cheat/SKILL.md:只把 episode 接到上游 cheat-on-content,不能保留本地教程型评分公式。教程与观点内容必须分开 calibration lane;state 迁移只允许先 dry-run。
  • laohan-yunying/SKILL.md:⑫—⑭的抖音运营编排器。发布准备要求 bianpai check --require final;真实发布仍必须 Jeffrey 确认。它保存 publish-record、数据快照和评论洞察,随后交给 upstream cheat-publish/retro;不自动发布、回复或改预测。
  • ①的 experiment 必须预先声明合法 metric_keys 和 T+N 窗口;⑬只有匹配该窗口、播放与全部预注册指标非空的 measurement_role: TARGET 快照可完成,其他观测只能标 CONTEXT,不得被 retro 当作假设结论。
  • bianpai vendors 写 schema 3 的 FROZEN_ON_RESUME vendor preflight,冻结 UP_TO_DATE|READY_LOCAL_AHEAD 状态和 Cheat/dbskill 完整 HEAD SHA。status/next 只做本地 HEAD 对照;任一 HEAD 漂移或旧 schema 2 均 BLOCKED,不能猜填或静默迁移。新 episode 必须走 sync-content-vendors.sh --new-episode 并创建 00-编排/executor-lock.json;registry/runtime 漂移时进行中 schema 2 episode BLOCKED,不得刷新 preflight 静默换执行器。
  • platform-display-evidence.json 是⑫可选的发布后展示核对;缺失不阻断,存在时必须绑定本期截图/导出、aweme、标题、封面 SHA、final SHA 和三项 MATCHED 人工结论。错绑记录必须使⑫未完成,绝不自动修正或伪造证据。
  • ④ score 后只算 PARTIAL,最终盲预测 RECORDED 后才算 COMPLETE;production/final/full 不得接受 PENDING。⑧媒体/SRT、⑨时间轴/PROOF、⑪ workspace/handoff/qa-evidence 都由 workflow 机械 gate 验证。
  • ③和⑤不能只凭当前稿 hash 通过:schema 2 的③还必须绑定未过期 ruleset review/expiry/scan 时间并写 platform_guarantee: false;⑤的 04-深扫报告.md 必须声明 review_status: CLEARunresolved_issue_count: 004-事实核验.md 必须声明 fact_check_status: CLEARcontradicted_count: 0unverifiable_count: 0。CONTENT_CLAIMS 在 AUTONOMOUS_RUN 不因通用 HIGH/CRITICAL 确认规则询问用户,而是回②最多三轮修订;不得伪 CLEAR。
  • schema 2 NATIVE 不再用 EDL visual.render_plan 决定审美:⑨先写 semantic contract/art direction/parity plan/native challenge,再确定性编译 animation-ast.json;⑪的两个 proposal/extension 必须绑定 AST/proposal SHA、完整 state/transition 与 capability allow-list。旧 render_plan 仅保留 PARITY 兼容回归,不能作为表现上限。
  • NATIVE 生产顺序必须是 proposal → prepare-renderer-extension → scaffold → native implementation → seal-renderer-workspace → base render → apply-review-captions → QA。未 seal workspace、renderer 内嵌字幕或缺 base/SRT/caption-style/candidate 哈希链都必须失败;Remotion/HyperFrames 的 base 都不得自行烧字幕。
  • ⑪ QA 的字幕安全区和有声审片结论必须来自独立 qa-reviews/<renderer>.json,绑定 candidate SHA、审阅者、时间及具体说明;任何命令行 PASS 都无效。每个 selected beat 均须有 entry/middle/exit 三帧证据。
  • ⑧的实际字幕必须经过 spoken-script-variance.json 与当前稿的事实差异审阅。任何事实偏离都回②→⑤→④,⑨不得自行解释或继续生产;仅 CLEAR、事实偏离为零的当前记录可放行。
  • bianpai vendors 成功后必须为本期写入绑定 runtime lock 的 00-编排/vendor-preflight.jsonstatus/next 缺少或检测到漂移时只能 BLOCKED,不能隐式联网或继续路由。
  • ⑤必须输出当前稿/事实报告绑定的 claim ledger。每个 PROOF beat 与 proof asset 都要绑定 SUPPORTED claim_id 和同源 evidence;只有 URL 或本地文件不能冒充该画面已证明口播事实。
  • 配套V5.1回归使用项目中真实存在的node scripts/test-v5-director-method.mjsnode scripts/test-codex-direct-contracts.mjs,并先运行node scripts/check-workflow-runtime.mjs;不得引用不存在的聚合脚本。修改被workflow-runtime-lock.json锁定的skill后必须同步更新内容SHA,否则bianpai必须BLOCKED。
  • dbskill 不再由 npx --all 更新:完整 vendor 在 ~/Documents/dbskill,并由 workflow 的 sync-content-vendors.sh fast-forward 后受管 symlink 到 Claude/Codex 通用 skills。视频 workflow 只固定调用 dbs-script-flow/dbs-resonate,条件调用 hook/ai-check/spread;其余能力见项目 docs/dbskill-编排映射.md
  • laohan-daoyan v2是V5导演预制的轻入口,不另建方法真源;实时config因registry/runtime漂移失败时必须报告阻断并停止,禁止手改lock。只有Jeffrey明确授权,才可由视频工作流的受控导演预制期迁移修复。V5仍为PILOT,真实本期candidate经Jeffrey验收前不得称为accepted或效果最优。⑩laohan-sucai按当前素材合同执行;laohan-donghua只保留METHOD_LAB历史兼容。
  • laohan-xiazai 的抖音下载按其 references/douyin.md:先用 opencli douyin user-videos 取得临时 play_url 并立即下载;opencli 失效才降级到移动端 UA + iesdouyin _ROUTER_DATA,再按文档后续降级。抖音不存在 YouTube/yt-dlp 路径。对标/cheat 批量文字转录用 MLX 未量化 large-v3;⑧停顿和词级时间真值用 whisper-timestamped large-v3。ASR-B01 证明 MLX 会合并静音间隔,不能替代⑧;不要使用会出现 NaN/吞吐异常的 Torch MPS。路径见 workflow CLAUDE.md 第24条。
  • schema 2 的①由 laohan-redian 唯一主写 signals/candidates/source-health 与 00-选题.*laohan-douyinsousuo 是可独立调用的 OpenCLI 只读取证 adapter,只写抖音证据,两者不合并、不做 programmatic skill-to-skill 调用。screening_summary 在同一个 candidates 文件内留下全部 signals→8—12 longlist(不足8个时全部)→短名单链路,不新增节点或第二套打分器。①至少比较两个候选,唯一 SELECTED 必须绑定 PRIMARY;平台可用时再绑定 PLATFORM_SIGNAL,平台预检/查询失败则 platform_alignmentcontent_gap 均可 UNAVAILABLE,不单独阻塞。当前 OpenCLI 1.8.6 的 douyin search 只有结果样本,hashtag hot 是全站热点,hashtag search 实测失败且 hot --keyword 不能可靠过滤,禁止反推搜索量、供给总量、增长或内容缺口。source health 仍覆盖 DISCOVERY|DOUYIN_SEARCH|PRIMARY_PROOF 并绑定本期结果 SHA;每个 metric key 都有正整数 T+N/DAY、方向、基线和抖音创作者中心来源。opencli hackernews top 是当前命令,禁止 hackernews hot。不得为①安装 TrendRadar/RSSHub/Obsei 或另一套浏览器/爬虫;外部项目只作设计参考。laohan-fengmianqiuzhi v1.12默认只走秋芝方向,从45套词典选择3个不同模板族,按推荐程度生成01—03三张9:16完整封面;不恢复排字脚本或评分流程。Jeffrey可改选,没有改选时发布准备默认以01为源候选。
  • laohan-chuangzuo 1.7.0 的新期②必须写schema 3并通过laohan-chuangzuo/scripts/check-script-contract.mjs。固定开场为“嘿,你有没有这种感觉,”。热点短小允许自然短稿;内容单位必须是“新判断+支撑+观众价值”,同义内容不得用于凑字数或时长。人味废话与内容重复分开,仍保留至少4种结巴、语气词、自嘲、反问或停顿设备;分层时使用连续1、2、3……。旧episode继续按executor lock中的1.6.x schema 2验证,不迁移既有稿件。
  • 上游 Cheat 的 prediction anatomy 现统一为所有 calibration 阶段都写 7 组件、bucket、概率、中枢与反事实;cold-start 只降低 Confidence、拉平分布。工作流 AUTONOMOUS_RUN 可用 Review Mode: AUTONOMOUS_AGENT_PROXY,但不得伪造 Jeffrey 确认或用户覆盖。旧 2026-07-12-native-method-real-run 已在拍摄前 SUPERSEDED;当前生产期 2026-07-16-codex-direct-production 已真实完成①(92条发现信号、30条OpenCLI抖音关键词证据、3个候选,选中C01),唯一下一步是②写稿。

Read the full file on GitHub · 48 lines

Changes

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.

  1. 9d ago First seen · 48 lines · 3,470 tokens per session scan A 2209f559fa36

Subscribe to this mod's changes

laohan-skills CLAUDE.md is an instructions file published in the GitHub repository hanzhcn/laohan-skills (11 stars, last pushed 13d ago), licensed MIT. It adds 3,470 tokens to every session, about $0.0174 per session on Opus 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 2026-08-30.

Related

Other instructions, from other repositories

next.js AGENTS.md

AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.

vercel/next.js · 7,296 tokens

codex AGENTS.md

AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.

openai/codex · 5,153 tokens

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

vscode oss-third-party-notices.instructions.md

Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).

microsoft/vscode · 5,001 tokens

langchain AGENTS.md

AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.

langchain-ai/langchain · 4,469 tokens

spec-kit AGENTS.md

AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.

github/spec-kit · 7,104 tokens