设置代理文件

文档 社区

技能解读

当前原文的自动摘译,尚未人工审校。

交互规则

只要存在当前代理的交互式询问/澄清工具,就使用它。提供简短的多项选择选项以及推荐的默认值,以便用户可以单击/选择而不是键入。如果当前代理没有询问工具,请在聊天中询问相同的问题并使用编号选项,并仅在用户回答后继续。不要默默地假设写操作的答案。一次最多询问一个设置决定。

1. 检查

首先查找已安装的项目技能:首选npx skills@latest list --json(如果有)。如果存在,请阅读skills-lock.json。仅在需要时扫描常用项目技能目录:.agents/skills/

2. 提议

总结:已安装的技能找到了现有的代理指导文件找到了值得创建或更新的支持文档然后在编写之前要求确认。对目标代理文件使用结构化选择,例如:

3. Agent指令块

在每个选定的代理文件中添加或更新一个块。保留周围的用户内容。使用此形状:如果文件已有 ## Agent skills 部分,则仅替换该部分。

4. 按已安装技能支持文档

仅创建对已安装技能有用的文档。

流狐档案 运行与权限由流狐静态检查推断,不代表作者声明或实际测试结果。
领域
文档
作者声明的智能体
未找到明确声明;不据此推断已兼容或已测试
静态检查
88 / 100 · 启发式扫描,不代表运行安全
运行环境证据
未执行;当前没有隔离运行测试证据
作者 / 版本 / 许可证
@tomevault-io · 未声明许可证
上下文用量评级
低消耗
接入复杂程度
即装即用
是否需要外部接口密钥
不需要
兼容的系统
macOS · Linux · Windows
运行环境要求
未检出明确要求
检测到的文件与系统行为
  • 只读
  • 允许写入 / 修改
网络行为
仅限本地
安装命令数
无(仅作为资料)

需要注意: 未声明工具权限范围;实际权限由运行环境决定。

作者说明文件(原文) SKILL.md

# Setup Agent Files

Create the repo guidance files that make installed skills usable across agents. Keep output minimal, explicit, and based on what is actually installed.

This is a prompt-driven setup skill: inspect, propose, ask with structured choices, then edit.

## Interaction rule

Use the current agent's interactive ask/clarification tool whenever it exists. Present short multiple-choice options with a recommended default so the user can click/select instead of typing.

If the current agent has no ask tool, ask the same question in chat with numbered choices and proceed only after the user answers. Do not silently assume answers for write operations.

Ask at most one setup decision at a time.

## 1. Inspect

Find installed project skills first:

- Prefer `npx skills@latest list --json` if available.
- Read `skills-lock.json` if present.
- Scan common project skill dirs only when needed:
  - `.agents/skills/`
  - `.claude/skills/`
  - `.codex/skills/`
  - `.gemini/skills/`
  - `.github/copilot/skills/`
  - `.cursor/skills/`
  - `skills/`

For each installed skill, read only its `SKILL.md` frontmatter first. Read the body only if the setup decision depends on it.

Also inspect existing guidance files:

- `AGENTS.md`
- `CLAUDE.md`
- `GEMINI.md`
- `.github/copilot-instructions.md`
- `.cursor/rules/*.mdc`
- `CONTEXT.md`, `CONTEXT-MAP.md`, `docs/adr/`
- `docs/agents/`

Do not assume missing files should be created.

## 2. Propose

Summarize:

- installed skills found
- existing agent guidance files found
- support docs worth creating or updating

Then ask for confirmation before writing. Use structured choices for the target agent files, for example:

- Existing agent files only (recommended when any exist)
- `AGENTS.md` only (recommended when none exist)
- `AGENTS.md` plus agent-specific files

If multiple agent files are possible, recommend:

- Update existing agent files first.
- If no agent file exists, create `AGENTS.md` as the shared default.
- Create agent-specific files only when the repo already uses that agent or the user asks for them.

## 3. Agent instruction block

Add or update one block in each selected agent file. Preserve surrounding user content.

Use this shape:

```markdown
## Agent skills

Installed skills live in the project skill directories. Before using a skill, read its `SKILL.md` and only load supporting files when needed.

### Installed skills

- `<skill-name>` — <one-line use case from frontmatter>

### Supporting docs

- `docs/agents/<file>.md` — <why this exists>
```

If the file already has an `## Agent skills` section, replace only that section.

## 4. Support docs by installed skill

Create only docs that are useful for the installed skills.

### Issue workflow docs

If any of these are installed: `to-issues`, `to-prd`, `triage`, `setup-agent-skills`, create or update:

- `docs/agents/issue-tracker.md`
- `docs/agents/triage-labels.md` when `triage` is installed

Ask where issues live with structured choices:

- GitHub Issues (recommended when a GitHub remote exists)
- GitLab Issues (recommended when a GitLab remote exists)
- Local markdown
- Other

Use existing remotes as evidence, not as final authority.

### Domain docs

If any of these are installed: `diagnose`, `tdd`, `grill-with-docs`, `improve-codebase-architecture`, `zoom-out`, create or update:

- `docs/agents/domain.md`

Ask whether to create domain docs with structured choices:

- Use existing `CONTEXT.md` / ADRs only
- Create or update `docs/agents/domain.md`
- Skip domain setup

Record how agents should consume `CONTEXT.md`, `CONTEXT-MAP.md`, and ADRs. Do not create domain docs unless the user confirms.

### Review docs

If any of these are installed: `review-swarm`, `review-and-simplify-changes`, create or update:

- `docs/agents/review.md`

Record review scope defaults, severity expectations, and whether review agents may edit files.

### Visual docs

If `visual-explainer` is installed, create or update:

- `docs/agents/visuals.md`

Record preferred output directory, whether generated HTML should be committed, and any sharing/deployment constraints.

### Repository reference docs

If `librarian` is installed, create or update:

- `docs/agents/repositories.md`

Record cache location preferences and rules for using remote repositories as references.

## 5. Keep it lean

Do not create docs for skills that do not need repo-specific configuration. Interaction/productivity skills such as `caveman`, `handoff`, `grill-me`, and `write-a-skill` usually only need to be listed in the agent instruction block.

Do not add local development notes, private setup details, or generated session context. If the repo has a private ignored directory for local notes, leave it ignored and do not reference its contents.

## 6. Done

Report:

- files created or updated
- installed skills covered
- any skills that need no repo-specific docs
- any decisions the user deferred

---
> Source: [0xNino/skills](https://github.com/0xNino/skills) — distributed by [TomeVault](https://tomevault.io).
<!-- tomevault:4.0:skill_md:2026-05-23 -->

讨论

基于 GitHub Discussions。登录 GitHub 即可参与讨论、点赞、订阅更新。