跳到正文
cccskills
免费在 GitHub 公开

td-feature-build

Implement ONE tdmcp tool from a spec — the canonical file pattern (Zod schema + …Impl + register… ToolRegistrar), a bridge Python payload via buildPayloadScript, and an offline msw unit test — touching only new files, never shared registries. Use when coding/implementing/writing a tdmcp tool, generator, effect, control, CHOP/TOP/SOP builder, or Python bridge work, especially when several features are built in parallel.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(1)

  • SKILL.md4.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

td-feature-build — implement one tool, new files only

Implement exactly what the spec says, in your own files, green in isolation. The parallel-build contract is simple: new files only; never edit anything shared (layer*/index.ts, tools/index.ts, cli/agent.ts, docs). That isolation is what lets builders run concurrently.

The canonical tool-file pattern

Every tool file exports a pure …Impl and a register…. Mirror a neighbour in the same layer, but the shape is:

import { z } from "zod";
import type { ToolRegistrar } from "../types.js";          // note the .js extension
import { runBuild, errorResult } from "../result.js";

export const fooSchema = z.object({
  name: z.string().default("foo"),
  intensity: z.number().min(0).max(1).default(0.5),
});

export async function fooImpl(ctx: ToolContext, args: z.infer<typeof fooSchema>) {
  // build via the client; never throw — return errorResult / runBuild on failure
}

export const registerFoo: ToolRegistrar = (server, ctx) =>
  server.registerTool(
    "create_foo",
    { title: "…", description: "…", inputSchema: fooSchema.shape },
    (args) => fooImpl(ctx, args),
  );

ctx is the ToolContext ({ client, knowledge, recipes, logger, vault?, allowRawPython }). …Impl is pure and unit-testable with a mocked client.

Hard rules

  • Never throw out of a handler. Validate via the Zod schema; turn TD failures into friendly isError results with errorResult / runBuild / friendlyTdError. Return, don't throw.
  • .js import extensions on every relative import (ESM/NodeNext). noUncheckedIndexedAccess is on — guard indexed access.
  • Biome: 2-space, double quotes, semicolons, trailing commas, 100-col. Check with ./node_modules/.bin/biome check <yourfiles> directly — npm run lint fails with a false ESLint parse error under the RTK proxy.
  • New files only. Your tool file + tests/unit/<feature>.test.ts. Do not touch shared files; report your export names to the integrator instead.

Bridge / Python work

  • Build the script with buildPayloadScript (encodes args as __PAYLOAD_B64__), execute through the client, parse the reply with parsePythonReport.
  • An executePythonScript payload must assign a result variable — the bridge reads it back.
  • If you edit td/ modules: keep TD-globals (op, app, project) inside functions so modules import cleanly, and run python3 -m py_compile on changed files.

TD gotchas that make a "successful" build actually cook

  • Many param/connect calls fail silently. Set the parameter that exists: a Level TOP has no gain — use brightness1. There is no ParMode.
  • No cross-container wires — route a signal out of a container through a Select TOP/CHOP, not a direct wire.
  • GLSL TOPs: declare out vec4 fragColor;; there is no built-in uTime (add your own uniform); avoid preamble #define collisions (F1/F2). KB shader snippets are references, not drop-in.
  • Reactive/time-dependent chains read 0 when the timeline is paused — that's expected, not a bug.

msw unit test

  • Mock the bridge with msw (no live TouchDesigner). Assert the real returned shape — operators created, params set, wiring, and isError paths — not a cast-away generic.
  • Run only yours: npx vitest run tests/unit/<feature>.test.ts. Green + biome-clean = done. You do NOT run the full build (the integrator does after wiring).

Output

The new tool file + test file, plus a note at _workspace/02_build_<feature>.md: files created, exported symbols (<feature>Impl, register<Feature>), target layer index, what the test asserts, and any spec deviation + reason. Then tell td-integrator the export names.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Use when manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling.

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

Use when a hybrid memory system that provides persistent, searchable knowledge management for AI agents (Architecture, Patterns, Decisions).

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

Reference for how BDB structures autonomous software engineering work — the seven-node dispatcher graph (Architect, TechLead, UI/UX, Engineering, Media/EventTech, Reviewer, Shipping) that /startcycle-graph actually runs. Use when you need the high-level lifecycle framing without inventing your own process.

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

Tools are how AI agents interact with the world. A well-designed tool is the difference between an agent that works and one that hallucinates, fails silently, or costs 10x more tokens than necessary. This skill covers tool design from schema to error handling.

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

Harness patterns for coding agents — memory, permissions, context engineering, delegation, skills, hooks, bootstrap.

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

agenttrail: live map of a multi-agent build in the browser: which plan component is being worked on, by which agent or harness, what is done and what is stuck. Use when a multi-agent pipeline starts (/startcycle, /startcycle-graph, /teamwork-preview) or after a plan-canvas approve, when the user asks to see what the agents are doing ("live map", "build board", "who is running now"), or by yourself whenever a multi-agent run is under way — no user prompt needed.

中文简介正在准备中。当前显示原文说明。

hybridlabor-api/aos62026年10月16日 更新

查看 hybridlabor-api 的全部技能

报告此技能的问题