AgentRC 指令生成

作者原文

Generate tailored AI agent instruction files via AgentRC instructions command. Produces .github/copilot-instructions.md (default, recommended for Copilot in VS Code) plus optional per-area .instructions.md files with applyTo globs for monorepos. Use after running /acreadiness-assess to close gaps in the AI Tooling pillar.

SkillPi 解读

能完成的任务

AgentRC 指令生成

适合谁

请按作者原文中的适用场景判断。

输入和输出

以作者完整文档为准；本站未运行验证产物。

软件、账号及付费要求

使用前按完整文档确认软件、账号、费用及权限；由宿主核对并授权实际文件读写和外部服务。文档引用域名（非运行白名单）：未列出 HTTP(S) 域名。 AgentRC 会读取仓库代码并使用所配置的外部模型；先确认代码可发送范围、模型提供商和费用，执行 dry-run，审阅生成指令后再确认覆盖。 外部工具按官方来源固定版本并单独核验；本站未安装或运行依赖。文件改动、外部发送、密钥和费用仍由使用者在宿主中逐项授权。

限制与区别

原文中的功能、版本状态和效果描述属于作者陈述；中文为对应翻译。仅核对固定版本及静态附件证据，不保证运行效果、外部服务或兼容性。

实测状态

未测试

安全评级与访问范围

caution · 静态证据核验，不代表运行安全保证。

使用前按完整文档确认软件、账号、费用及权限；由宿主核对并授权实际文件读写和外部服务。文档引用域名（非运行白名单）：未列出 HTTP(S) 域名。 AgentRC 会读取仓库代码并使用所配置的外部模型；先确认代码可发送范围、模型提供商和费用，执行 dry-run，审阅生成指令后再确认覆盖。 外部工具按官方来源固定版本并单独核验；本站未安装或运行依赖。文件改动、外部发送、密钥和费用仍由使用者在宿主中逐项授权。

使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地；本站的核验不会授予任何运行权限。

原始 SKILL.md

---
name: acreadiness-generate-instructions
description: 'Generate tailored AI agent instruction files via AgentRC instructions command. Produces .github/copilot-instructions.md (default, recommended for Copilot in VS Code) plus optional per-area .instructions.md files with applyTo globs for monorepos. Use after running /acreadiness-assess to close gaps in the AI Tooling pillar.'
argument-hint: "[--output .github/copilot-instructions.md|AGENTS.md] [--strategy flat|nested] [--areas | --area <name>] [--apply-to <glob>] [--claude-md] [--dry-run]"
---

# /acreadiness-generate-instructions — write AI agent instructions

Use this skill whenever the user wants to **create**, **regenerate**, or **refresh** their custom instructions for AI coding agents (Copilot, Claude, etc.). This is the *Generate* step in AgentRC's **Measure → Generate → Maintain** loop and the single highest-leverage action for the **AI Tooling** pillar.

## Output options

VS Code recognises several instruction file types — AgentRC generates the most common ones:

| File | Scope | When to use |
|---|---|---|
| `.github/copilot-instructions.md` | Always-on, whole workspace | **Default** — VS Code Copilot's native instruction file |
| `AGENTS.md` | Always-on, whole workspace | Multi-agent repos (Copilot + Claude + others) |
| `.github/instructions/*.instructions.md` | Scoped by `applyTo` glob | Per-area / per-language rules in monorepos |
| `CLAUDE.md` | Claude-specific | Add via `--claude-md` (nested only) |

## Strategies

- **`flat`** *(default)* — single `.github/copilot-instructions.md` at the chosen path. Simple, easy to review.
- **`nested`** — hub at `.github/copilot-instructions.md` + per-topic detail files at `.github/instructions/<topic>.instructions.md`, each with an `applyTo` glob so VS Code only loads the topic when it's relevant. Better for large or multi-stack repos.

> **Why `.github/instructions/` and not `.agents/`?** AgentRC's default nested layout writes to `.agents/`, which is the right home for *agent-agnostic* repos (Copilot + Claude + Cursor reading `AGENTS.md`). For VS Code Copilot specifically, the native location is `.github/instructions/` with `applyTo` frontmatter — that's what Copilot auto-discovers. This skill rewrites AgentRC's nested output to the VS Code-native location whenever the main output is `.github/copilot-instructions.md`. If you instead chose `--output AGENTS.md`, nested keeps AgentRC's default `.agents/` layout.

For monorepos, generate **area-scoped** instructions with `--areas`, `--area <name>`, or `--areas-only`. Areas are defined in `agentrc.config.json`. Per-area output is written as VS Code `.instructions.md` files with an `applyTo` glob (see below).

### Topic vs area `.instructions.md` files

Both end up in `.github/instructions/` but they answer different questions:

| Kind | Filename example | `applyTo` example | Where it comes from |
|---|---|---|---|
| **Topic** (nested) | `testing.instructions.md` | `**/*.{test,spec}.{ts,tsx,js}` | AgentRC `--strategy nested` topic split |
| **Area** (monorepo) | `frontend.instructions.md` | `apps/frontend/**` | `agentrc.config.json` areas + `--areas` |

You can have both at once: a nested set of topic files plus per-area files for a monorepo.

## Per-area files with `applyTo`

When the user opts into areas, emit one VS Code-native `.instructions.md` file per area at `.github/instructions/<area>.instructions.md`. Each file MUST start with frontmatter declaring the glob the rules apply to:

```markdown
---
applyTo: "apps/frontend/**"
---

# Frontend area instructions

…AgentRC-generated content for this area…
```

Workflow:

1. **Read `agentrc.config.json`** to discover declared areas and their `paths` / globs. If `paths` is missing, ask the user for the glob (e.g. `src/api/**`).
2. **Run `agentrc instructions --areas`** (or `--area <name>`) to produce the per-area body content.
3. **Wrap each area's content** in `.github/instructions/<area>.instructions.md` with the `applyTo` frontmatter taken from the area's `paths`. If the user passed `--apply-to <glob>` on a single-area call, use that glob verbatim.
4. **Leave the main file alone** — the root `.github/copilot-instructions.md` stays as the always-on instructions; `.instructions.md` files only kick in for matching paths.

Naming: lowercase, kebab-case area name. Examples: `.github/instructions/frontend.instructions.md`, `.github/instructions/api.instructions.md`, `.github/instructions/infra.instructions.md`.

## Steps

1. **Pick the target file**. **Default to `.github/copilot-instructions.md`.** Switch to `AGENTS.md` only if the user mentions multi-agent / Claude / Cursor support.
2. **Always ask which strategy to use** — `flat` or `nested` — unless the user already specified one in their message or via `--strategy`. Present the trade-off briefly:
   - **Flat** *(default)* — one `.github/copilot-instructions.md`. Simple, easy to review in a single PR. Best for small/medium repos with one stack.
   - **Nested** — hub `.github/copilot-instructions.md` + per-topic `.github/instructions/<topic>.instructions.md` files (each with an `applyTo` glob so VS Code only loads them when relevant). Best for large or multi-stack repos. Add `--claude-md` to also emit `CLAUDE.md`.
   Recommend `nested` proactively when the repo has > 5 top-level directories, multiple stacks, or already uses a monorepo tool (turbo/nx/pnpm workspaces).
3. **Detect monorepo areas** by reading `agentrc.config.json`. If areas exist, ask the user whether they want **per-area `.instructions.md` files with `applyTo`** in addition to the root file. Default to "yes" when `agentrc.config.json` declares areas.
4. **Run dry-run first** so the user can preview:
   ```bash
   npx -y github:microsoft/agentrc instructions --output <file> --strategy <flat|nested> [--areas|--area <name>] [--claude-md] --dry-run
   ```
5. **Show a short summary** of what would change — files that would be created or overwritten, area count + their `applyTo` globs, model used (default `claude-sonnet-4.6`).
6. **On confirmation, run the same command without `--dry-run`** (and optionally `--force` if files already exist).
7. **Post-process layout for Copilot output**:
   - **If `--output` ends in `copilot-instructions.md` and strategy is `nested`**: move/rewrite AgentRC's `.agents/<topic>.md` files to `.github/instructions/<topic>.instructions.md`. Add frontmatter to each file with an appropriate `applyTo` glob (see "Topic applyTo defaults" below). Delete the now-empty `.agents/` directory.
   - **If `--areas` was used**: also write `.github/instructions/<area>.instructions.md` for every area, using each area's `paths` from `agentrc.config.json` as the `applyTo` glob (override with `--apply-to` for single-area calls).
   - **If `--output AGENTS.md`** was chosen: keep AgentRC's native `.agents/` layout for nested — agent-agnostic readers expect it there.
   Create the `.github/instructions/` directory if missing.

### Topic `applyTo` defaults

When promoting AgentRC's nested topic files to `.instructions.md`, use these defaults unless the user specifies otherwise:

| Topic | Default `applyTo` |
|---|---|
| `testing` | `**/*.{test,spec}.{ts,tsx,js,jsx,mjs,cjs}` |
| `style` / `code-quality` / `formatting` | `**/*.{ts,tsx,js,jsx,mjs,cjs,py,go,rs,java,kt,cs}` |
| `build` / `ci` | `**/{package.json,turbo.json,nx.json,.github/workflows/**}` |
| `docs` | `**/*.md` |
| `security` | `**` |
| anything else / hub-level | `**` |
8. **Verify** by reading the generated file(s) back and showing the user a 1-paragraph synopsis: stack detected, conventions captured, length, list of `.instructions.md` files with their globs.
9. **Suggest next steps**:
   - Re-run the `assess` skill to confirm the AI Tooling pillar score improved.
   - If the user already has both `copilot-instructions.md` and `AGENTS.md`, recommend consolidating to a single source of truth (AgentRC flags this at maturity Level 2+).

## Notes

- AgentRC reads your **actual code** — no templates. Output reflects detected languages, frameworks, and conventions.
- `--claude-md` (nested strategy only) also emits `CLAUDE.md`.
- VS Code applies `.instructions.md` files automatically when the active file matches `applyTo`. The root `.github/copilot-instructions.md` always loads.
- Never run this skill non-interactively in CI; instructions are part of the repo and should land via PR.

版权与许可声明

MIT License

Copyright GitHub, Inc.

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.