MCP Server Builder
anthropics/skills
Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.
为 Henji-AI 新增、修改或迁移应用能力,并完成内置 Pi 与现代 MCP 的同源适配。新增工作区、页面、浮层或工具入口、用户数据、设置、业务操作、长任务、稳定引用、权限、宿主上下文、能力搜索,或清理旧 HostCommand/HostQuery/Agent 工具时使用;纯样式、布局、文案和不改变业务能力的组件调整不使用本 skill。
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install henjicc/Henji-AI henji-application-capability --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.codex/skills/henji-application-capability .claude/skills/henji-application-capability && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .claude/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capabilityType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install henjicc/Henji-AI henji-application-capability --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.codex/skills/henji-application-capability .agents/skills/henji-application-capability && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .agents/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install henjicc/Henji-AI henji-application-capability --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.codex/skills/henji-application-capability .cursor/skills/henji-application-capability && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .cursor/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/henjicc/Henji-AI.git --path .codex/skills/henji-application-capability--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install henjicc/Henji-AI henji-application-capability --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.codex/skills/henji-application-capability .gemini/skills/henji-application-capability && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .gemini/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install henjicc/Henji-AI henji-application-capabilityInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .github/skills && cp -r skills-src/.codex/skills/henji-application-capability .github/skills/henji-application-capability && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .github/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add henjicc/Henji-AI --skill henji-application-capability -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install henjicc/Henji-AI henji-application-capability --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/henjicc/Henji-AI.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.codex/skills/henji-application-capability .opencode/skills/henji-application-capability && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "henji-application-capability" agent skill from https://github.com/henjicc/Henji-AI/tree/main/.codex/skills/henji-application-capability into .opencode/skills/henji-application-capability/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "henji-application-capability", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
henji-application-capability为 Henji-AI 新增、修改或迁移应用能力,并完成内置 Pi 与现代 MCP 的同源适配。新增工作区、页面、浮层或工具入口、用户数据、设置、业务操作、长任务、稳定引用、权限、宿主上下文、能力搜索,或清理旧 HostCommand/HostQuery/Agent 工具时使用;纯样式、布局、文案和不改变业务能力的组件调整不使用本 skill。
Henji Application Capability is an agent skill from henjicc/Henji-AI. 为 Henji-AI 新增、修改或迁移应用能力,并完成内置 Pi 与现代 MCP 的同源适配。新增工作区、页面、浮层或工具入口、用户数据、设置、业务操作、长任务、稳定引用、权限、宿主上下文、能力搜索,或清理旧 HostCommand/HostQuery/Agent 工具时使用;纯样式、布局、文案和不改变业务能力的组件调整不使用本 skill。
Its SKILL.md is about 2.8k tokens, which your agent loads only when the skill is triggered. The skill folder holds 4 other files, including reference files (for example `agents/openai.yaml` and `references/capability-patterns.md`).
It works with Model Context Protocol. The repository describes itself as: 痕迹AI - 一个软件用上各种AI!聚合多家供应商,一站式生成图片、视频和音频. The licence is Apache-2.0.
8 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit 1019cf2. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Shell commands in SKILL.md call:
npmFrom the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md. Its commands use npm, which can reach the network depending on how they are called.
From URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Henji Application Capability loads about 2.8k tokens when it runs, and up to ~4.2k if it reads all its reference files. Until then it costs about 50 tokens; SKILL.md has 543 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from henjicc/Henji-AI at commit 1019cf2, republished under its Apache-2.0 licence (© henjicc). 543 words, ~2,754 tokens.
.claude/skills/henji-application-capability/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.UI、内置 Pi 与现代 MCP 共用正式领域服务和应用运行实例。公共契约位于 src/core/application-control/,渲染层在 src/features/application-control/ 装配各领域,Electron 的 services/application-runtime/ 协调授权、执行、账本、费用与恢复,先于 Pi 和 MCP 初始化。
Pi 与 MCP 只负责工具和结果投影,不拥有业务状态。MCP 固定使用正式协议 2026-07-28 与拆分 SDK 2.0.0,拒绝旧协议;授权逐请求核验,持久操作不随网络连接结束。旧自研助手、Henji Script、发现租约和旧前端工具桥已移除,不保留兼容执行链或旧数据迁移要求。
将领域能力接入公共注册和执行入口。先找正式业务服务,再声明或扩展实体;不要为 Pi 或 MCP 复制业务实现。
先回答下面七个问题,再决定工作深度:
按答案分三级执行:
| 级别 | 判断 | 本 skill 的动作 |
|---|---|---|
| 仅呈现变化 | 前六项均否,第七项为是 | 停止本 skill;改用 henji-ui-surface。不得新增能力、实体、schema 或 MCP 工具 |
| 覆盖核对 | 只新增页面、浮层、入口或导航;数据和操作完全复用现有正式能力 | 核对 Surface/导航登记、已有实体与能力的可发现性、入口是否直接复用正式服务。记录核对结论,不新增同义实体、专用能力或 MCP schema |
| 能力增量 | 前五项任一为是,或现有正式声明无法表达新入口的业务语义 | 继续完整流程:先找唯一业务实现,再补领域声明、执行器、权限、账本、结果验证与多入口投影 |
多个 skill 可以同时命中,但职责不能重复:henji-ui-surface 管界面层级与呈现,canvas-node-builder 管画布节点结构,henji-model-adaptation 管供应商与模型契约,本 skill 只管应用能力增量和覆盖证明。不要为满足 skill 数量制造额外适配层。
默认不写。 反射层已经提供三个通用动词,领域只要注册实体和属性,助手就能用:
| 需求 | 做法 |
|---|---|
| 读某个状态 | 注册实体和属性即可,list_application_entities / read_application_entity 自动可用 |
| 改某个已有对象的属性 | 注册属性并实现 ApplicationMutationExecutor,change_application_entities 自动可用 |
| 新增或删除集合成员 | 实体描述里声明 collectionWrite,实现 ApplicationCollectionExecutor |
| 带算法的语义操作 | 才写专用 ApplicationCapabilityDefinition |
静态属性与 collectionWrite 只表示“结构上支持”。每个 provider 还必须通过 getPropertyAvailability / getCollectionAvailability 返回当前引用、模式和状态下的真实可用性;没有额外集合限制时复用 unrestrictedCollectionAvailability。计划、提交预检和每步执行前由统一事务引擎复核,领域服务里不再复制模式守卫。
describe_application_entities 可带 refs 查询实例级动态状态。recovery 只放结构化能力/实体/属性标识;任何操作步骤必须先通过正式 describe → change → read/真相源 结果测试,禁止凭读代码猜一条路线写进提示或错误。
只有当动作无法用属性写入表达时才写专用能力——例如"环绕运镜"要按角度采样算轨迹,"复用或布置对象"要做碰撞检测和复用判定。凡是"设置某某值""加一条记录"这类,一律走通用动词。
Camera Stage 只公开 camera_stage.state_keyframe。动画操作必须走已经过结果测试的路径:在同一次 change_application_entities 的 changes 中按顺序交替写 camera_stage.playback.current_time 与对象、角色或摄像机的 animatable.* / pose_preset,最后可在同一事务写 loop/playing;应用会自动创建或更新完整场景状态,任一步失败则整体回滚。除非后一步需要前一步新创建且尚未返回的引用,否则禁止按时间点拆成多轮调用。派生属性轨道只供播放与导出,不能注册为公开实体或持久化真相源。
多项实体写入使用同一通用事务,算法操作经正式领域能力执行。输入约束由反射和能力 schema 投影;跨领域使用完整稳定引用,不创建额外脚本编排入口。
声明了可写属性就必须注册 ApplicationMutationExecutor,声明了 collectionWrite 就必须注册 ApplicationCollectionExecutor。每个实体必须至少拥有一种写入执行器,或填写 writeExclusion.reason,明确说明为何只读以及状态由哪个正式模块或操作维护;不得用“暂时不支持”代替判断。三者由覆盖测试强制一致。
新增可写属性只在一处声明。 一个属性此前要碰 4 个位置(属性描述符、读取映射、写入表项、界面动作账本),缺任何一处都是静默失效——不报错,助手安静地少一块能力,只有用户实机撞上才发现(三维场景外观 24 项当初就是这样漏掉描述符和读取两处)。现在统一走 src/core/application-control/fieldDefinition.ts 的 ApplicationFieldDefinition:
sceneField('sky_color', '天空颜色', COLOR, {
read: (settings) => settings.sky.color,
write: (store, value) => store.setSceneSkyColor(value),
storeAction: 'setSceneSkyColor',
})一条声明用 fieldDescriptors() / fieldReadValues() / fieldWriterTable() / fieldLedgerEntries() 派生出描述符、读取映射、写入表项、账本条目四样东西,四个消费方各取所需。字段定义按领域收在 <领域>Fields.ts(如 cameraStageSceneFields.ts、canvasFields.ts、assetFields.ts),领域内部再包一层 <领域>Field() 薄封装填好该领域固定的 entityType、权限、revision scope。同一个 store 动作被多个字段共用时(如 updateObject 一次改 name/visible/color/character_variant 四个属性),fieldLedgerEntries() 按声明顺序把它们累进同一条账本绑定。禁止再分别手写这四处——统一定义之后漏一条是整条从四处一起消失,会被 storeActionCoverage 门禁当场抓到,而不是像以前那样只漏两处却全绿。
同一份领域声明现在同时服务三个入口:应用界面、内置 Pi、外部智能体(MCP)。新增能力时只写一次声明,三个入口各自投影;任何一处出现第二份业务字段表都算缺陷。
| 要让外部智能体看到的东西 | 唯一声明处 | 谁来投影 |
|---|---|---|
| 工具名、参数、必填项 | ApplicationCapabilityDefinition 的 Zod inputSchema | localApplicationHost.ts 用 z.toJSONSchema 投影,随宿主注册送到主进程 |
| 哪些域/实体可读可写 | 反射注册表的 exposures、requiredPermissions.write、collectionWrite | src/features/application-control/externalCapabilityInventory.ts |
| 有意只读及其原因 | 实体的 writeExclusion.reason | 同上,投影成外部契约里的 readOnlyReason |
| 通用读改增删的公开写入范围 | 同上派生结果 | 随注册跨进程送达,operationCoordinator 直接消费 |
| 写入的并发与幂等信封 | 协议层固定的 operationId + baselineIds | electron/main/services/application-runtime/toolCatalog.ts 统一注入 |
由此得到几条硬要求:
electron/main/services/mcp/** 写任何业务字段、实体类型或属性清单。 那里只允许协议层自身的参数(操作标识、分块偏移、契约发现),业务参数一律从能力定义投影。前缀白名单尤其禁止——它和领域声明是两份真相,新增写域时必然漂移。resolveOperationTargets,创建操作还要用 resolveOperationWriteTargets 绑定操作身份;纯导航由注册表统一绑定视图身份。external.kind: delegate 和实际实体、操作及属性,覆盖检查必须证明它们确实开放且可写。内部协议写明 internal 原因,保存恢复写明 recovery 并仅走原操作账本。不得用排除掩盖缺失的业务执行器。applicationControlCoverage.test.ts 核对全部前端能力的 MCP 路由、执行器、目标绑定和委托字段,配合属性、集合、Surface 与 store 动作门禁。新增功能只留在组件里而不登记不算完成;必须加入现有正式注册入口。内置 Pi 默认延迟加载新领域工具,不需要跟着新增一份启用名单。writeExclusion.reason。它是"未完成"和"有意排除"在机器上唯一的区分方式,并且会原样出现在外部契约里,成为调用方改道的依据。不接受"暂时不需要"这类无法验证的表述。守这几条的门禁:electron/main/services/application-runtime/toolCatalog.test.ts(schema 同源与授权过滤)、src/features/application-control/externalCapabilityInventory.test.ts(八个业务写域与只读原因)、collectionCoverage.test.ts(每个实体要么能写要么写明原因)。
src/core/application-control/domains/ 的领域能力模块中声明 ApplicationCapabilityDefinition 并注册到统一目录,由公共应用入口调用。additionalProperties: false,禁止开放 patch、storePatch、executeScript、script 等任意 Store Patch 或代码入口。不得新增源码执行字段。ApplicationRef;禁止传原始密钥、本地路径或不受控的大对象。详细字段选择和范式见 references/capability-patterns.md。
kind: 'command'、kind: 'query'、HostCommand 或 HostQuery 分支。start、completed、failed,日志只记录稳定引用和脱敏信息。availability 和前置条件中明确声明。MCP 按授权返回稳定排序的声明目录,宿主未就绪时不隐藏静态工具;执行给出准确状态。Pi 按需披露同一目录,不能另写业务名单。
实体与属性约束只从真实注册表派生,模型猜测和当前页面不能隐藏已注册业务能力。
执行结果以真实输出、副作用、领域回读和持久化状态为依据;模型说明失败不能改写已经发生的业务事实。
幂等键按调用者隔离,执行前登记;未知结果先查询。取消等待不取消生成,保存恢复不重新执行业务动作。
为能力提供用户可能使用的中文、英文和领域别名。
声明 acceptsRefs、producesRefs 和前置依赖,让跨模块任务通过稳定引用衔接。
Router 只提供页面锚点和搜索建议,不得以分类结果限制能力可用性或授权。
写能力必须通过 control.impacts 声明 Effect、实体和属性;一次输出可能影响多个目标时实现 resolveObservedEffects(input, output),从真实结果解析数量、稳定引用和验证证据。没有解析器的能力一次最多贡献一个 Effect。
跨领域结果传递必须注册接收上游稳定引用的正式桥梁能力,并验证正式下游调用成功;禁止让模型猜测领域等价物、手工拼接内部路径,或把生成结果冒充素材。媒体 URL/本地路径只在宿主内部组合服务中流动。
反射层公开每个可写属性真实接受的 writeOperations;高层集合 set 由计划器确定性编译为 append/remove 最小差异,不支持的操作在计划期拒绝。
同一通用事务可以重复写同一属性,最终状态只验证最后一次写入;播放头、播放开关等会话控制声明 verificationStrategy: 'execution',中间状态的真实领域副作用必须由正式结果测试覆盖。
后置步骤依赖前序步骤刚建立的动态可用状态时,静态权限仍在计划期拒绝,动态 availability 延迟到该步骤执行前复核,失败由事务补偿。
entityType 与 target.kind 重复表达实体类型时由通用适配器统一规范化;领域 provider 只可在全局唯一时补全短引用,歧义引用继续拒绝。
写入触发自动创建/更新等领域级联副作用时,正式执行器必须返回带静态 declarationId 的强类型 Effect Receipt;evidence 只做验证与说明,禁止用它记账或只按输入猜影响范围。
拒绝必须能被自我修正:实体类型写错就列出该域注册了哪些,属性写错就列出这个实体有哪些,参数被静默丢弃就说清丢了哪些键与可用的是哪些。只给错误码等于逼模型继续猜,而它猜不中就是死循环。
AI 可见输入 schema 不暴露 baseRevision / expectedRevisions。并发基线只由 Gateway expected-revision 信封传入,不得形成第二条 revision 路径。
优先用通用动词覆盖(见第 0 节);只有算法型语义操作才注册专用能力。
新增或修改下列对象时,注册能力或加入带原因的显式排除清单:
check:assistant-capabilities 或覆盖测试阻断。不得因为“暂时没有助手需求”而省略覆盖判断。
resolveSurfaceObservationProfile 判断一次,目录和覆盖清单都从它派生;界面标注 Surface ID 时从目录反查,不在组件里复制映射表。target="window" 整窗,任何时候都可用;只在需要聚焦时传具体 surfaceId。截图范围永远只有本应用窗口,禁止桌面和其他应用窗口。data-observation-sensitive。新增界面时,凡是把明文本地路径、密钥或令牌渲染出来的节点都要自己标上;type="password" 的输入框自带圆点掩码,不需要标。observe_application_surface 是应用截图入口,原始媒体通过正式媒体读取入口取得。不要新增只返回媒体引用或“已截图”标记的观察能力——模型看不到画面却会以为看过了;要产生视觉证据就返回 verificationKind: 'visual_pending_model' 加合法附件。readEntity 断言世界真的变化,completed 或 evidence 不算结果断言。docs/rules/testing.md 确定最小匹配检查。涉及公共核心契约及关键依赖的整轮重构在最终集成做一次 L3,不在每次搬移后重复全量。test:reality 隔离场景,检查实际截图和日志;产物过期只需 electron:bundle。不得以浏览器代替 Electron。能做的事优先通过通用动词暴露;专用能力只用于无法用属性写入表达的算法型操作。
声明 collectionWrite 的实体类型都注册了 ApplicationCollectionExecutor,由覆盖测试拦截。
每个 provider 都实现动态集合可用性;模式限制可在 describe 阶段看见,并由事务引擎统一执行。
每个实体都有 mutation/collection 执行器或非敷衍的 writeExclusion.reason;新增实体后运行 npm run check:assistant-capabilities。
Application Control 反射注册表是实体、属性和集合 CRUD 的唯一元数据源;ApplicationCapabilityDefinition 是算法操作的唯一元数据源;Pi 与 MCP 工具目录都只能投影两者,不能成为第三份手写 schema。
新增已登记的业务实体或属性后,外部智能体无需任何 MCP 侧改动即可读写;有意只读的实体带得住 writeExclusion.reason。
对外输入、输出和错误通过公共契约校验,并由固定现代协议的官方 Client 经 HTTP 验证。
正式业务服务是唯一业务执行源。
普通界面不显示开发性解释。
新代码没有旧 command/query 兼容路径。
新代码没有任意 Store Patch、任意脚本执行或 Application API 核心跨层导入。
权限、revision、日志、引用和成功证据均有自动化验证。
用户编辑 A 时可后台读改、生成并保存 B;打开 B 复用同一实例和历史,关闭页面不结束已提交工作。
保存失败保留脏状态;删除和退出经过实例屏障,只有已保存、无任务、无使用者时才释放。
© henjicc, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 2 other files (references) in .codex/skills/henji-application-capability of henjicc/Henji-AI.
Open the folder on GitHubat commit 1019cf2
Henji Application Capability next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Henji Application Capability this skillhenjicc/Henji-AI | 254 | — | ~2.8k | Automated safety check: Pass | Apache-2.0 | |
| MCP Server Builderanthropics/skills | 180k | 63 repos | ~2.3k | Automated safety check: Pass | Apache-2.0 | |
| MCP Server BuildershareAI-lab/learn-claude-code | 78k | 5 repos | ~1.2k | Automated safety check: Pass | MIT | |
| MCP Integration for Pluginsanthropics/claude-plugins-official | 38k | 11 repos | ~3.1k | Automated safety check: Pass | Apache-2.0 | |
| Figma use_figma Plugin API Ruleswarpdotdev/warp | 65k | 4 repos | ~4.4k | Automated safety check: Pass | AGPL-3.0 | |
| Stitch to Remotion Walkthrough Videosgoogle-labs-code/stitch-skills | 8.4k | 6 repos | ~3.2k | Automated safety check: Notes | Apache-2.0 |
anthropics/skills
Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.
shareAI-lab/learn-claude-code
Walks through building MCP servers in Python or TypeScript that expose tools, resources and prompts to Claude, with templates, registration and testing.
anthropics/claude-plugins-official
Explains how to bundle Model Context Protocol servers in a Claude Code plugin, covering config files, stdio, SSE, HTTP and WebSocket server types, and authentication.
warpdotdev/warp
Required groundwork before any use_figma call: the rules and reference files for running JavaScript in a Figma file through the Plugin API without common failures.
google-labs-code/stitch-skills
Builds walkthrough videos from Stitch design projects using Remotion, with transitions, zoom effects and text overlays on each screen.
coollabsio/coolify
A skill your agent uses for Laravel MCP development. An agent skill from coollabsio/coolify.
henjicc/Henji-AI
Henji-AI 画布(ReactFlow)新增/改造节点时使用。指导如何选择普通生成节点、节点内工具工作台、标准行拼装或纯展示节点,如何声明 CanvasNodeDefinition,以及如何让遮罩、打光、镜头等特殊交互各自不同但保持统一骨架。普通模型 schema 会被标准生成节点自动读取;只改供应商、模型参数、显隐、计价或请求构建时不使用本 skill。
henjicc/Henji-AI
Henji-AI 新建或改造任何界面/页面骨架/面板/弹窗/侧栏/设置分区/节点 UI,或调整按钮层级、分隔线、颜色、图标、毛玻璃、动画、层级时使用。主文件涵盖“石墨”设计系统速览(主题引擎与语义令牌、按钮五档默认静默、淡强调底选中态、尺寸档位)、页面骨架的横向条带上限与命令带、表面层级(surface/elevation)铁律、五级容器词汇表、分隔线准入、选项集合静息态与选中态词汇表、必须复用…
henjicc/Henji-AI
面向 Henji-AI 的模型与供应商调研、文档整理、参数体验和适配工作流。用于“新增供应商”“给现有供应商新增模型”“模型和供应商都要新增”“核查 API/价格/平台别名”“校对参数顺序、通用交互、隐藏参数或默认请求值”这类需求;先输出确认清单,用户确认后再实施。普通模型 schema 会被标准画布节点自动读取,不因模型会出现在画布里而自动触发节点开发。
henjicc/Henji-AI
在痕迹AI修图、调色、抠出或选中主体、移除物体、修补瑕疵、编辑图层与蒙版、导出图片或流转图片产物时使用。视频时间线与成片用 video-edit-workbench;写代码画面用 video-edit-code-creation。
henjicc/Henji-AI
在剪辑里设计、编写或修改代码素材(动态图形、标题、花字、模板、滤镜),统一风格、编排动画、放入时间线并取帧审查时使用。普通剪辑、媒体生成和进度查询不触发。
henjicc/Henji-AI
在痕迹AI剪辑工作台做粗剪、整理时间线、节奏、字幕、配乐与音量、响度、调色、转场、多机位、竖版改画幅、导出,或把生成、画布、图片文档、口播产物放进剪辑及把剪辑帧送到其他工作区时使用。写代码画面用 video-edit-code-creation;单独写代码素材不触发本技能。
Works with
为 Henji-AI 新增、修改或迁移应用能力,并完成内置 Pi 与现代 MCP 的同源适配。新增工作区、页面、浮层或工具入口、用户数据、设置、业务操作、长任务、稳定引用、权限、宿主上下文、能力搜索,或清理旧 HostCommand/HostQuery/Agent 工具时使用;纯样式、布局、文案和不改变业务能力的组件调整不使用本 skill。. Henji Application Capability is an agent skill from henjicc/Henji-AI.
Run `npx skills add henjicc/Henji-AI --skill henji-application-capability -a claude-code`. Or copy the skill folder (.codex/skills/henji-application-capability in henjicc/Henji-AI) into .claude/skills/henji-application-capability in your project. Claude Code loads it when a task matches its description.
Run `npx skills add henjicc/Henji-AI --skill henji-application-capability -a codex`. Or copy the skill folder (.codex/skills/henji-application-capability in henjicc/Henji-AI) into .agents/skills/henji-application-capability in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add henjicc/Henji-AI --skill henji-application-capability -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/henji-application-capability, .gemini/skills/henji-application-capability, .github/skills/henji-application-capability and .opencode/skills/henji-application-capability in your project.
Going by SKILL.md and its folder, Henji Application Capability needs the command-line tools its instructions call (npm).
SKILL.md contains no URLs. Its commands use npm, which can reach the network depending on how they are called. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
Henji Application Capability is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 2.8k tokens (SKILL.md is roughly 11k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 1.5k tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Henji Application Capability: MCP Server Builder (anthropics/skills, 180k stars), MCP Server Builder (shareAI-lab/learn-claude-code, 78k stars), MCP Integration for Plugins (anthropics/claude-plugins-official, 38k stars) and Figma use_figma Plugin API Rules (warpdotdev/warp, 65k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
henjicc (a GitHub user) maintains it in henjicc/Henji-AI, which has 254 GitHub stars. The repository holds 11 skills in this directory. The repository was last updated on October 9, 2026.
Source: henjicc/Henji-AI on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.