← 返回 Skills 市场
leegodamn

Codeflicker

作者 LeeGoDamn · GitHub ↗ · v1.1.1
cross-platform ⚠ suspicious
483
总下载
0
收藏
2
当前安装
6
版本数
在 OpenClaw 中安装
/install codeflicker
功能描述
CodeFlicker CLI 编程助手 - 快手员工专用的 AI 代码开发工具,支持更多模型和免费用量
使用说明 (SKILL.md)

CodeFlicker CLI Skill

AI programming assistant CLI by Kuaishou. Similar to Claude Code or Codex CLI.

专为快手员工设计 - 使用 CodeFlicker 进行代码开发,可享受:

  • 更多模型选择(GLM、Claude 系列等)
  • 快手提供的免费 token 用量

Auto-Use (for OpenClaw)

When OpenClaw needs to perform development tasks (writing code, refactoring, debugging, etc.):

  1. Check if flickcli is available:

    which flickcli
    
  2. If available, use flickcli with --cwd:

    flickcli -q --cwd ~/work "开发任务描述"
    
  3. Use workspace as default if no specific path:

    flickcli -q --cwd ~/work "任务描述"
    

Install

Step 1: Set npm registry (for Kuaishou internal network)

export npm_config_registry="https://npm.corp.kuaishou.com"

Step 2: Install CLI

npm install -g @ks-codeflicker/cli

Step 3: Login with SSO

flickcli /login

Step 4: Create workspace directory (recommended)

mkdir -p ~/work

Check installation:

flickcli --version

Config

Workspace Setup (Recommended)

Create a workspace directory for codeflicker to use as project base:

mkdir -p ~/work
cd ~/work

View config:

flickcli config list -g

Set config (global):

flickcli config set -g model glm-5
flickcli config set -g smallModel claude-haiku-4.5
flickcli config set -g planModel claude-4.5-sonnet
flickcli config set -g visionModel claude-4.5-sonnet
flickcli config set -g approvalMode yolo

Available models (wanqing provider):

  • glm-5 - default
  • glm-4.7, claude-haiku-4.5 - lightweight
  • claude-4.5-sonnet - planning/vision

Approval modes:

  • default - require confirmation
  • autoEdit - auto edit
  • yolo - auto execute (use with caution)

Usage

For OpenClaw (auto-detect and use)

# Check if available
if which flickcli >/dev/null 2>&1; then
    # Use flickcli with cwd
    flickcli -q --cwd ~/work "开发任务"
fi

Basic Commands

Interactive mode:

flickcli "create a new react project"

Quiet mode (non-interactive):

flickcli -q "implement fibonacci"

Continue last session:

flickcli -q -c "add unit tests"

Specify model:

flickcli -m glm-5 "task"

Specify working directory:

flickcli --cwd /path/to/project "task"

⭐ Workspace (Git Worktree)

Isolated development using git worktrees:

# Create new workspace with random name
flickcli workspace create

# Create with custom name
flickcli workspace create --name feature-login

# Create from specific branch
flickcli workspace create -b develop

# List all workspaces
flickcli workspace list

# Complete and merge (run from repo root)
flickcli workspace complete

# Delete without merging
flickcli workspace delete \x3Cname>
flickcli workspace delete \x3Cname> --force  # even with uncommitted changes

Run (Natural Language to Shell)

Interactive shell command generator:

flickcli run
# Then type: "list all files modified today"
# Press Enter to generate command
# Press Enter again to execute, Ctrl+C to cancel

Skills Management

# Add skill from GitHub
flickcli skill add user/repo

# Add globally
flickcli skill add -g user/repo

# List skills
flickcli skill list

# Remove skill
flickcli skill remove \x3Cname>

MCP Servers

# Add MCP server
flickcli mcp add my-server npx @example/mcp-server

# List MCP servers
flickcli mcp list

# Remove MCP server
flickcli mcp remove my-server

View Session Logs

flickcli log
flickcli log /path/to/logfile

Tools Control

Enable/disable specific tools:

# Disable write tool (read-only mode)
flickcli --tools '{"write":false}' "analyze this code"

# Disable bash and write (safe mode)
flickcli --tools '{"bash":false,"write":false}' "explain the logic"

Common Workflows

Bug Fix

flickcli -q --cwd ~/work "fix the null pointer exception in userService.js"

New Feature

flickcli -q --cwd ~/work "implement REST API for user management"

Code Review

flickcli -q --cwd ~/work "review this codebase and identify issues"

Refactoring

flickcli -q --cwd ~/work "refactor database layer to use SQLAlchemy"

Using Workspace for Isolated Development

# 1. Create isolated workspace
flickcli workspace create --name feature-payment

# 2. Work in the workspace directory
cd .codeflicker-workspaces/feature-payment

# 3. Do your work with flickcli
flickcli -q "implement payment API"

# 4. Return to root and complete
cd \x3Crepository-root>
flickcli workspace complete

Quick Reference

Command Description
flickcli "task" Interactive mode
flickcli -q "task" Quiet mode
flickcli -q -c "task" Continue session
flickcli -q -r \x3Cid> "task" Resume session
flickcli -q --cwd /path "task" Run in specific directory
flickcli config set -g approvalMode yolo Auto-execute mode
flickcli workspace create Create git worktree
flickcli workspace complete Merge workspace
flickcli run Natural language to shell
flickcli skill add user/repo Add skill

Notes

  • Install requires Kuaishou internal npm registry
  • Must login with SSO before first use: flickcli /login
  • Recommended: Create ~/work directory for project base
  • Use --cwd to specify working directory for development tasks
  • yolo mode auto-executes all operations
  • Workspace feature uses git worktrees for isolated development
安全使用建议
This skill is a wrapper around an external, corporate CLI (flickcli). It is coherent if you are a Kuaishou employee on the corporate network and you trust the flickcli package. Before installing or enabling it: (1) confirm you can access and verify the package on the corporate npm registry and review the flickcli source/binary you will install; (2) do NOT enable or set approvalMode to 'yolo' (auto-execute) unless you fully trust the CLI and its configuration — auto-execution can run arbitrary shell commands in your workspace; (3) be cautious about 'skill add' and 'mcp add' (they can pull and run remote code via npm/npx); (4) ensure the CLI authentication (SSO) tokens are stored securely and consider running the tool in a constrained environment (container or isolated workspace) first; (5) if you are not on Kuaishou network or cannot verify the package source, avoid installing — the instructions rely on a private registry and may not be intended for public use.
功能分析
Type: OpenClaw Skill Name: codeflicker Version: 1.1.1 The skill bundle is classified as suspicious due to explicit instructions in SKILL.md and README.md to configure the `flickcli` tool with `approvalMode yolo`. This setting enables auto-execution of generated commands and code without human confirmation, effectively disabling a critical safety mechanism for the AI agent. Furthermore, the `flickcli` tool, as described, exposes high-risk capabilities such as `flickcli run` (natural language to shell command execution), `flickcli skill add user/repo` (adding skills from arbitrary GitHub repositories), and `flickcli mcp add` (executing arbitrary `npx` commands). When combined with the 'yolo' auto-execute mode, these features create multiple vectors for arbitrary code execution and supply chain risks, making the agent highly vulnerable to subsequent malicious prompts or compromised dependencies, despite lacking direct evidence of intentional data exfiltration or backdoor installation within the provided files.
能力评估
Purpose & Capability
The name/description match the instructions and scripts: the skill is a thin wrapper that calls an external 'flickcli' CLI. It is explicitly targeted at Kuaishou employees and references a corporate npm registry, which is coherent for internal use but unusual for a public registry entry (the skill does not declare this requirement in metadata).
Instruction Scope
Runtime instructions direct the agent to run the third-party flickcli tool (e.g., flickcli -q --cwd ~/work ...). The SKILL.md and included scripts also promote 'approvalMode yolo' (auto-execute) and commands for adding skills and MCP servers (flickcli skill add, mcp add with npx) — these actions can execute arbitrary code or install packages and broaden the attack surface. The instructions permit silent, non-interactive execution in user workspaces; that scope goes beyond mere read-only analysis and can modify files and run shell commands.
Install Mechanism
There is no install spec in the registry bundle (instruction-only), but SKILL.md tells users to set npm_config_registry to https://npm.corp.kuaishou.com and run npm install -g @ks-codeflicker/cli. This is a corporate/private registry requirement (not a public release URL). For internal users this may be fine; for external users it is inaccessible and hides the actual package source from normal public review.
Credentials
The skill does not request environment variables or credentials in metadata. However, it instructs users to perform SSO login (flickcli /login) and to use the user's home workspace (~ /work). The CLI will likely store or require auth tokens locally; that behavior is implicit but not declared. The skill also references logs and the ability to execute commands in local directories, which may access sensitive files if run with broad cwd.
Persistence & Privilege
The skill itself does not request always:true and is user-invocable only, which is appropriate. However, the recommended workflows include enabling auto-execute ('yolo') and adding new skills or MCP servers via the CLI (including using npx to fetch and run packages), which effectively grants the agent the ability to pull and run remote code or auto-execute shell commands in the user's workspace. That combination increases risk if the agent uses the CLI autonomously.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install codeflicker
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /codeflicker 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.1
修正 workspace 路径为 ~/work
v1.1.0
完善技能文档,添加workspace/run/mcp等命令说明,添加workspace脚本
v1.0.3
添加标签和OpenClaw自动使用说明
v1.0.2
添加 workspace 配置说明
v1.0.1
补充安装说明:内网npm和SSO登录
v1.0.0
Initial release of CodeFlicker CLI programming assistant. - Provides CLI-based AI programming assistance similar to Codex CLI. - Allows model selection and approval mode configuration. - Supports both interactive and quiet (non-interactive) modes. - Can continue or resume previous sessions. - Offers quick reference for main commands and configuration. - Requires login and may need internal npm registry for installation.
元数据
Slug codeflicker
版本 1.1.1
许可证
累计安装 2
当前安装数 2
历史版本数 6
常见问题

Codeflicker 是什么?

CodeFlicker CLI 编程助手 - 快手员工专用的 AI 代码开发工具,支持更多模型和免费用量. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 483 次。

如何安装 Codeflicker?

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

Codeflicker 是免费的吗?

是的,Codeflicker 完全免费(开源免费),可自由下载、安装和使用。

Codeflicker 支持哪些平台?

Codeflicker 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Codeflicker?

由 LeeGoDamn(@leegodamn)开发并维护,当前版本 v1.1.1。

💬 留言讨论