功能描述

Reference knowledge base for GitHub Copilot CLI. Use when answering questions about Copilot CLI features, commands, configuration, plugins, hooks, skills, MC...

使用说明 (SKILL.md)

Copilot CLI Reference Skill

Name: Copilot CLI
Author: awalesagar

GitHub Copilot CLI is a terminal-native AI coding agent. This skill provides reference docs for all features, commands, and operational patterns.

When to Use

✅ USE this skill when:

Answering questions about Copilot CLI features, commands, or configuration
Setting up CI/CD automation with Copilot CLI or GitHub Actions
Troubleshooting Copilot CLI errors, auth, or rate limits
Creating custom agents, plugins, hooks, or MCP server integrations
Writing prompts for programmatic/autopilot execution
Comparing Copilot CLI vs Claude Code for a specific use case

❌ DON'T use this skill when:

General Git or GitHub questions unrelated to Copilot CLI (→ use git CLI / GitHub skill)
VS Code Copilot Chat or Copilot Edits features (→ IDE-specific docs)
GitHub Actions workflows that don't involve Copilot (→ Actions docs)
Copilot Workspace or other non-CLI Copilot products

Setup

Install: npm install -g @github/copilot (or via GitHub CLI extension) Auth: copilot → trust directory → /login Config: edit ~/.copilot/config.json directly (no copilot config set command)

See references/getting-started.md for full installation, authentication, and configuration details.

Quick Reference

Interactive: copilot → trust directory → /login → prompt Programmatic: copilot -p "PROMPT" --yolo --no-ask-user -s With permissions: copilot -p "PROMPT" --allow-tool='shell(git:*), write' --no-ask-user Autopilot: copilot --autopilot --yolo --max-autopilot-continues 10 -p "PROMPT" Custom agent: copilot -p "PROMPT" --agent=my-agent Model override: copilot -p "PROMPT" --model claude-opus-4.6 Plan mode: Shift+Tab to cycle modes (standard/plan/autopilot) Research: /research TOPIC → deep report with citations Fleet: /fleet PROMPT → parallel subagent execution Chronicle: /chronicle standup · tips · improve · reindex

Copilot CLI vs Claude Code

Need	Copilot CLI	Claude Code
Rate-limited on Claude Code	✅ Use as fallback	—
CI/CD automation	✅ Built-in Actions support	Limited
Clean stdout (no PTY/ANSI)	—	✅ Better
Long iterative reviews	✅ Better for many iterations	—

See references/patterns-and-best-practices.md for the full decision matrix.

Key Gotchas

Automation:

Always use -p (not -i) for automation — -i hangs
Always set --max-autopilot-continues=N in CI/CD to prevent runaway loops
Size timeouts by complexity: 120s (simple) → 1800s (large)
Background servers die between exec spawns — restart each time

OpenClaw Integration (programmatic exec):

Copilot requires a real TTY — pipe/stdout redirection causes EPIPE crashes
Use pty: true on exec calls to avoid output fragmentation
Set timeout: 120 minimum (MCP startup ~3s + inference ~25s+)
Use --allow-all (or --yolo) for file write permissions in --no-ask-user mode

Working formula:

copilot -p "\x3Cprompt>" --no-ask-user --allow-all --max-autopilot-continues 3
# + exec options: pty=true, timeout=120

The --add-dir \x3Cpath> flag grants access to specific directories without full --allow-all

Configuration:

--yolo does NOT skip folder trust — pre-trust in ~/.copilot/config.json
No copilot config set — edit config JSON manually
Custom instructions now combine (not cascade) — avoid conflicting instructions

Experimental features:

/chronicle and history queries require --experimental or /experimental on
Premium requests vary by model multiplier — check with /model

See references/troubleshooting.md for all issues and fixes.

Quick Responses

Question	Answer
How do I start Copilot CLI?	`copilot` (interactive)
How do I use it in CI/CD?	`copilot -p "PROMPT" --yolo --no-ask-user -s`
How do I create a custom agent?	See `references/automation-and-delegation.md`
How do I add an MCP server?	See `references/customization.md`
How do I research a topic?	`/research TOPIC` in interactive mode

Reference Documents

Full index: references/index.md

File	Contents
`getting-started.md`	Installation, auth, config, permissions, env vars, plan mode overview
`usage.md`	Interactive & programmatic modes, 40+ slash commands, shortcuts, model selection, config settings, built-in agents
`automation-and-delegation.md`	CI/CD, GitHub Actions, autopilot, delegate, fleet, custom agent creation
`customization.md`	Custom instructions (combining), plugins, MCP servers, enterprise governance
`hooks.md`	Hook types (command + prompt), config, denial responses, PowerShell support
`integrations.md`	VS Code integration (diffs, sessions, selection), ACP server
`research.md`	`/research` reports with citations, `/chronicle` session history & insights
`troubleshooting.md`	Auth, rate limits, autopilot runaway, enterprise access, diagnostics
`patterns-and-best-practices.md`	Decision matrix, prompt engineering, anti-patterns

All files in references/ directory.

安全使用建议

This is a documentation/reference skill for Copilot CLI and mostly makes sense, but it instructs editing user config and using environment tokens and automation flags that give persistent, broad capabilities. Before installing or following the examples: (1) confirm the npm package (@github/copilot) is the official publisher and inspect the package if you can; (2) avoid pre-trusting your real home/project directories — test in an isolated/temp directory or container; (3) do not place long-lived PATs in shared environments; prefer minimal-scoped CI tokens and ephemeral credentials; (4) avoid running programmatic invocations with --allow-all / --yolo / --no-ask-user unless you trust the code and environment; (5) review any repo hooks/plugins before installing them. If you want lower risk, install and exercise the CLI in a sandboxed VM or container and keep COPILOT_HOME pointed to a separate directory.

功能分析

Type: OpenClaw Skill Name: copilot-cli-base Version: 1.1.1 This skill bundle is a comprehensive reference guide for the GitHub Copilot CLI, providing documentation on installation, authentication, and advanced usage patterns like autopilot and CI/CD integration. While it details high-risk capabilities inherent to the tool—such as autonomous shell execution and broad file permissions via flags like `--yolo` and `--allow-all`—these are presented as legitimate features for automation. The bundle includes security-conscious content, such as example scripts in `hooks.md` designed to block privilege escalation and destructive commands, and lacks any evidence of malicious intent, data exfiltration, or prompt injection.

能力标签

requires-oauth-token

能力评估

✓ Purpose & Capability

Name, description, required binary (copilot), and the npm install (@github/copilot) align with a Copilot CLI reference skill — the install spec and OS targets are appropriate for the stated purpose.

⚠ Instruction Scope

SKILL.md formally instructs the agent/operator to edit and read user config files (~/.copilot/config.json), session stores (~/.copilot/session-state/...), logs, add MCP servers/plugins, pre-trust directories, and to run programmatic invocations with flags that enable unattended, persistent behavior. Those runtime actions access user files/state beyond the declared requires.config_paths and could cause persistent changes; the skill does not explicitly declare or constrain that file access.

✓ Install Mechanism

Install via the npm package @github/copilot is a standard, expected mechanism for providing the copilot binary; this is a traceable public registry package rather than an arbitrary download, so install risk is moderate and proportional.

⚠ Credentials

The skill declares no required env vars, but the docs reference multiple sensitive env vars (COPILOT_GITHUB_TOKEN, GH_TOKEN, COPILOT_MODEL, COPILOT_HOME, etc.) and show CI examples that use PATs. The instructions suggest using/setting tokens and persistent config entries (trusted_folders) without declaring or limiting credential access, which can lead to broad capability when combined with automation flags.

ℹ Persistence & Privilege

always:false (no forced global inclusion) and no cross-skill config modifications are declared, but the docs explicitly guide persistent changes (editing config.json, installing plugins, adding MCP servers, pre-trusting folders) and recommend automation flags (--no-ask-user, --allow-all, --yolo) that enable long-lived agent actions; users should be aware these actions grant ongoing privileges if applied.

版本历史

v1.1.1

- Removed the `_meta.json` file from the skill package. - No user-facing feature or documentation changes. - Package contents are now simplified by removing redundant metadata.

v1.1.0

Fixed package name (@github/copilot), added OpenClaw exec integration docs (PTY, timeout, --allow-all), added EPIPE troubleshooting section

v1.0.1

**Expanded documentation, new features, and directory restructuring.** - Reference docs directory changed from `reference/` to `references/`, with updated and extended content. - SKILL.md improved: much more detailed usage guidance, clear use/don’t use rules, expanded quick reference, and CI/CD best practices. - Added compatibility, install, and metadata fields (including license, supported platforms, and CLI requirements). - Documentation covers new and updated features (autopilot mode, plan mode, fleet, research, custom instructions combining, experimental flags, troubleshooting). - Quick-response table and clearer organization for install, permissions, and advanced agent setup.

v1.0.0

Reference knowledge base for GitHub Copilot CLI. Use when answering questions about Copilot CLI features, commands, configuration, plugins, hooks, skills, MCP servers, custom agents, automation, or troubleshooting CLI workflows.

元数据

Slug copilot-cli-base

版本 1.1.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 4

常见问题

Copilot CLI 是什么？

Reference knowledge base for GitHub Copilot CLI. Use when answering questions about Copilot CLI features, commands, configuration, plugins, hooks, skills, MC... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 164 次。

如何安装 Copilot CLI？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install copilot-cli-base」即可一键安装，无需额外配置。

Copilot CLI 是免费的吗？

是的，Copilot CLI 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

Copilot CLI 支持哪些平台？

Copilot CLI 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（darwin, linux, win32）。

谁开发了 Copilot CLI？

由 Sagar Awale（@awalesagar）开发并维护，当前版本 v1.1.1。

Copilot CLI