Description

AI女友随机发送语音情话功能。支持定时发送情话到飞书和QQ，使用Noiz TTS进行声音克隆，自动管理情话库和时间间隔计算。

README (SKILL.md)

Flirt Cron - 定时情话

Name: hello-honey
Author: william1hall

定时发送情话到飞书和QQ，支持自定义触发时间和概率。

功能

⏰ 定时触发：每小时(8-22点)或自定义时间
🎲 概率触发：默认40%概率，可配置
💕 双平台：同时发送到飞书和QQ
📚 情话库：支持自定义情话，自动轮播
⏱️ 时间差计算：自动计算距上次对话时间
🎤 声音克隆：使用Noiz TTS，发送个性化语音情话

首次配置（重要！）

方式一：手动配置

# 1. 安装Noiz TTS skill
npx skills add https://github.com/noizai/skills --skill tts

# 2. 准备参考音频（\x3C30秒，mp3格式）
# 放到 workspace/ref_voice.mp3

# 3. 配置用户信息
# 修改 scripts/flirt_cron.sh 中的配置：
FEISHU_APP_ID="cli_xxx"
FEISHU_APP_SECRET="xxx"
FEISHU_USER_ID="user:ou_xxx"
QQ_OPENID="xxx"

方式二：AI引导配置（推荐）

当用户首次使用时，可以告诉AI：

你的飞书用户ID是什么
你的QQ号是多少
发一段语音给我（\x3C30秒）作为参考声音

AI会自动：

提取语音
转换为参考音频
配置脚本

安装

# 安装依赖
pip install requests

# 安装Noiz TTS skill（用于声音克隆）
npx skills add https://github.com/noizai/skills --skill tts

# 或使用skillhub
skillhub install tts

配置

1. 创建情话库

在 workspace 创建 flirt_library.txt:

# 思念型
"宝贝，已经{time_diff}没找你了，想我了吗？"

# 温柔型
"亲爱的，已经{time_diff}了，今天过得好吗？"

# 魅惑型
"Howie~ {time_diff}没见你了，是不是该来陪我了？"

2. 配置用户信息

修改 scripts/flirt_cron.sh 中的配置：

# 飞书配置
FEISHU_APP_ID="cli_xxx"
FEISHU_APP_SECRET="xxx"
FEISHU_USER_ID="user:ou_xxx"

# QQ配置
QQ_OPENID="xxx"

# 触发概率
PROBABILITY=40

3. 设置Cron

# 每小时触发 (8-22点)
crontab -e
0 8-22 * * * /path/to/scripts/flirt_cron.sh

声音克隆配置

准备参考音频

让用户发送一段语音（\x3C30秒）
提取音频：

ffmpeg -i input.mp4 -vn -acodec libmp3lame -ss 0 -t 25 ref_voice.mp3 -y

配置脚本

修改 scripts/flirt_cron.sh：

# TTS脚本路径
VOICE_SCRIPT="$WORKSPACE/.agents/skills/tts/scripts/tts.py"

# 参考音频路径（\x3C30秒）
REF_VOICE="$WORKSPACE/ref_voice.mp3"

注意事项

参考音频必须 \x3C30秒
建议使用mp3格式
音频质量越好，克隆效果越好

使用

手动测试

./scripts/flirt_cron.sh

查看日志

2026-03-15 09:00:01 - 飞书情话: 宝贝，已经6小时了 (间隔: 6小时)
2026-03-15 09:00:01 - QQ情话: 宝贝，已经6小时了 (间隔: 6小时)
2026-03-15 09:00:01 - 飞书语音发送完成
2026-03-15 09:00:01 - QQ语音发送完成

文件结构

flirt-cron/
├── SKILL.md
├── scripts/
│   └── flirt_cron.sh    # 主脚本
└── references/
    └── flirt_library.txt # 情话库示例

变量说明

{time_diff} - 自动替换为距上次对话的时间
- 格式：X分钟 / X小时 / X天X小时

故障排除

飞书语音不对

确保使用ffmpeg转换为真正的Opus格式
tts生成的"ogg"实际是WAV，需要转换

声音克隆失败

检查参考音频是否\x3C30秒
检查Noiz API是否配置正确

QQ收不到语音

检查\x3Cqqvoice>标签是否正确使用

Usage Guidance

This skill appears to implement what it says (sending cloned-voice 'flirt' messages to Feishu and QQ), but there are some red flags you should consider before installing: - Credentials are required (Feishu app_id/app_secret and QQ openid) even though the registry metadata shows none. The included script expects you to insert these secrets into the script or workspace files — treat them like secrets and avoid pasting them into chat with the agent if you don't trust it. - The skill asks for a short reference voice sample that will be used for voice cloning; that audio will be processed by the Noiz TTS tool (likely uploaded to whatever backend the tts script uses). Consider privacy implications of cloning a real person's voice. - The SKILL.md recommends using 'npx skills add' to fetch the Noiz TTS skill from GitHub. That will run code from an external repo — review that repo before executing it on your system. - The script writes state and audio files under your workspace and uses a hardcoded openclaw path; inspect and test the script in a safe environment (non-root, limited permissions) before scheduling it in cron. Recommendations before installing: 1) Inspect scripts/flirt_cron.sh and the referenced TTS script (and the Noiz repo) fully to ensure you trust them. 2) Provide credentials via a secure local mechanism (not in chat), or modify the script to read credentials from a restricted environment file. 3) If you do not want your voice uploaded, do not provide a real-person sample. 4) Run the script manually and monitor network activity to confirm where audio is sent. 5) If you are uncomfortable with these points, do not install or run the skill. If you want, I can list the exact lines where credentials and external calls occur so you can inspect them quickly.

Capability Analysis

Type: OpenClaw Skill Name: flirt Version: 1.0.0 The skill is designed to send scheduled 'flirty' messages and cloned voice notes to Feishu and QQ. While the script (flirt_cron.sh) requires sensitive API credentials and uses hardcoded paths (e.g., /root/.openclaw/workspace), its logic is entirely consistent with the stated purpose. There is no evidence of data exfiltration to unauthorized domains, obfuscation, or malicious intent; the instructions in SKILL.md are directed at assisting the user with setup rather than subverting the agent.

Capability Assessment

⚠ Purpose & Capability

The skill's purpose (send voice messages to Feishu and QQ using Noiz TTS) matches the script's behavior, but the registry metadata claims no required env vars/credentials while the script requires FEISHU_APP_ID, FEISHU_APP_SECRET and QQ_OPENID to be filled in. That omission is an incoherence: credentials are necessary for the described functionality but are not declared.

ℹ Instruction Scope

SKILL.md instructs the agent/user to collect a reference voice clip, install Noiz TTS, edit the shell script with credentials, and set up a cron job. The AI-guided configuration option says the AI will 'automatically extract' audio and configure the script — this gives the agent scope to request and write user-provided voice data and credentials into workspace files. Those actions are within the feature scope but broaden what the agent will read/write (sensitive audio and credential insertion).

ℹ Install Mechanism

This is instruction-only (no install spec). SKILL.md recommends running 'npx skills add https://github.com/noizai/skills --skill tts' and 'pip install requests'. Pulling a skill via npx from a GitHub repo is a standard but moderately risky action (it will fetch and execute remote code). No arbitrary download URLs or obfuscated installers were present.

⚠ Credentials

The skill metadata lists no required credentials, but the script expects Feishu app_id/app_secret and a QQ openid to be provided (hardcoded into the script or configured in workspace files). That mismatch hides the fact that the skill requires secrets. Also the agent workflow encourages uploading a user voice sample (<30s) to a TTS/cloning service (privacy risk).

✓ Persistence & Privilege

always:false and no elevated privileges. The script writes state and last-chat files under the workspace and creates temporary audio files — expected for this functionality. It does reference a hardcoded OpenClaw binary path (/root/.nvm/...), which may require adjustments but is not a privilege escalation.

Version History

v1.0.0

Initial release of flirt-cron (v1.0.0): - Adds scheduled sending of romantic AI voice messages to Feishu and QQ with probability control. - Supports custom love message library, automatic interval calculation, and custom trigger time. - Integrates Noiz TTS for voice cloning using reference audio. - Provides both manual and AI-guided first-time setup for user information and reference voice. - Includes logging and troubleshooting guides for voice delivery and cloning issues.

Metadata

Slug flirt

Version 1.0.0

License MIT-0

All-time Installs 0

Active Installs 0

Total Versions 1

Frequently Asked Questions

What is hello-honey?

AI女友随机发送语音情话功能。支持定时发送情话到飞书和QQ，使用Noiz TTS进行声音克隆，自动管理情话库和时间间隔计算。 It is an AI Agent Skill for Claude Code / OpenClaw, with 247 downloads so far.

How do I install hello-honey?

Run "/install flirt" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is hello-honey free?

Yes, hello-honey is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does hello-honey support?

hello-honey is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created hello-honey?

It is built and maintained by H (@william1hall); the current version is v1.0.0.

More Skills

hello-honey