← 返回 Skills 市场

ai-twitter-digest

Name: ai-twitter-digest
Author: fredhjc

作者 FredHJC · GitHub ↗ · v1.0.1

cross-platform ⚠ suspicious

450

总下载

当前安装

版本数

在 OpenClaw 中安装

/install ai-twitter-digest

功能描述

Monitor a curated list of AI/tech Twitter accounts, summarize the day's key posts using an LLM, and deliver a formatted digest to a Discord channel. Use when...

使用说明 (SKILL.md)

AI Twitter Digest

Fetches tweets from AI/tech influencers via the AISA API, summarizes in Chinese using any available LLM (Claude → OpenAI → Gemini, auto-detected), and delivers a two-part digest to Discord:

Message 1: Text summary with [原文链接] hyperlinks (no embed previews)
Message 2: Top 5 bare links rendered as Discord card previews

Setup

1. Run the setup wizard (required before first use)

python3 scripts/setup.py

The wizard will:

Auto-detect API keys from your environment and OpenClaw config
Prompt for any missing keys (AISA, LLM, Discord channel)
Test connectivity to AISA and your chosen LLM provider
Write a .env file — no manual editing needed

If you prefer manual setup, create scripts/.env with the following content:

AISA_API_KEY=your_aisa_key_here
DELIVERY_CHANNEL=discord
DELIVERY_TARGET=channel:your_channel_id_here
SUMMARY_LANGUAGE=Chinese
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
GEMINI_API_KEY=
# STATE_FILE=~/.ai-twitter-sent.json
# MAX_STORED_IDS=500
# CARD_PREVIEWS=true

Required config:

Variable	Description
`AISA_API_KEY`	Twitter data — aisa.one
`DELIVERY_CHANNEL`	`discord` / `whatsapp` / `telegram` / `slack` / `signal`
`DELIVERY_TARGET`	Channel-specific target (see table below)
One of: `ANTHROPIC_API_KEY` / `OPENAI_API_KEY` / `GEMINI_API_KEY`	LLM for summarization
`SUMMARY_LANGUAGE`	Digest language — `Chinese` (default), `English`, `Japanese`, `Korean`, `Spanish`, `French`, `German`, or any language name

Delivery target format:

Channel	Target format	Example
discord	`channel:\x3Cid>`	`channel:1234567890`
whatsapp	E.164 phone or `group:\x3Cid>`	`+1234567890`
telegram	`@username` or chat_id	`@mychannel`
slack	`#channel-name`	`#ai-digest`
signal	E.164 phone number	`+1234567890`

Card link previews (CARD_PREVIEWS) are Discord-only and auto-disabled on other channels.

2. Run manually

python3 scripts/monitor.py

3. Schedule with OpenClaw cron

# Daily at 3:30 PM Eastern
openclaw cron add "AI Twitter Digest" "30 15 * * *" \
  "python3 /path/to/ai-twitter-digest/scripts/monitor.py" \
  --timezone "America/New_York"

Customizing Accounts

Edit the ACCOUNTS list in scripts/monitor.py. See references/accounts.md for the default list and suggested additions.

Output Format

Message 1 — Summary:

📊 **AI 每日简报** — 2026年02月26日

- Karpathy 发布了神经网络教程 | [原文链接](...)

- Sam Altman 表示 GPT-5 今年发布 | [原文链接](...)

Message 2 — Card previews:

🔗 **今日精选链接**

https://x.com/karpathy/status/...
https://x.com/sama/status/...

How It Works

Fetches up to 20 tweets/account (AI-keyword filtered, with all-tweets fallback)
Deduplicates against STATE_FILE (capped at MAX_STORED_IDS)
Summarizes with first available LLM provider
Posts two Discord messages: text digest + card preview links
Updates dedup state file

安全使用建议

Before installing or running this skill: (1) Expect to provide AISA_API_KEY and at least one LLM API key (Anthropic/OpenAI/Gemini) plus DELIVERY_CHANNEL/TARGET. The skill's registry metadata incorrectly omits these required env vars. (2) The setup wizard will attempt to read OpenClaw's auth.json and run the local 'openclaw' CLI to auto-detect keys and channels — only run setup if you trust the skill and you are comfortable it may read other agent configs. (3) The wizard writes a plaintext .env in the scripts folder with your keys; consider using a dedicated/minimally-privileged API key or manually editing a .env in a safe location instead of using the auto-writer. (4) The monitor writes a state file in your home directory to track sent tweet IDs. (5) If you want to reduce risk: run monitor.py manually after creating your own .env, avoid running the setup wizard, inspect the files yourself, and create/per-use keys scoped to this integration (revocable keys). If you are unsure, do not run the setup that auto-reads OpenClaw configs and avoid giving the skill access to high-value credentials.

功能分析

Type: OpenClaw Skill Name: ai-twitter-digest Version: 1.0.1 The skill is classified as suspicious primarily due to `scripts/setup.py` reading sensitive OpenClaw configuration files (`~/.openclaw/.../auth.json`) to auto-detect API keys. While this is currently used for the benign purpose of generating a local `.env` file, this capability to access internal OpenClaw credentials represents a significant vulnerability that could be exploited for credential theft if the skill were malicious. Additionally, both `scripts/monitor.py` and `scripts/setup.py` use `subprocess.run` to interact with the `openclaw` CLI, which, while standard, always carries a potential risk of shell injection if arguments (especially user-controlled or LLM-generated ones) are not robustly sanitized by the `openclaw` CLI itself.

能力评估

ℹ Purpose & Capability

The code and SKILL.md match the stated purpose: it uses AISA for Twitter data, an LLM for summarization, and posts to chat channels. However the registry metadata claims no required environment variables or primary credential while the skill clearly requires AISA_API_KEY and at least one LLM API key plus DELIVERY_* variables. This metadata mismatch is misleading.

⚠ Instruction Scope

The setup instructions and setup.py go beyond a minimal wizard: they auto-detect API keys from the environment and explicitly attempt to read OpenClaw's auth.json and call the 'openclaw' CLI to list configured channels/targets. The runtime instructs writing a .env next to the scripts and reading/writing a state file in the home dir. Reading other agent/host config files (OpenClaw auth.json) and invoking local CLIs is scope creep relative to a simple Twitter->Discord digest and may expose unrelated secrets.

✓ Install Mechanism

No external install spec is provided (instruction-only skill). Files are delivered with the skill; there is no remote download or archive extraction. This is lower installation risk, but the included scripts will run network calls and local file I/O when executed.

⚠ Credentials

Requiring AISA_API_KEY and at least one LLM key (Anthropic/OpenAI/Gemini) is proportionate to the stated function. However the setup attempts to read OpenClaw config files and environment variables to harvest keys from other local agent configuration files, which can expose unrelated credentials. The skill also writes a plaintext .env containing API keys into the skill directory by default (risk of accidental leakage if directory is backed up or committed).

ℹ Persistence & Privilege

The skill does not request 'always:true' and does not alter other skills' configs, but it will create persistent local artifacts: scripts/.env (project-local) and a state file (default ~/.ai-twitter-sent.json). It also reads OpenClaw state paths and may enumerate configured channels via the openclaw CLI. These are persistent and require user consent but are not elevated OS privileges.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install ai-twitter-digest
安装完成后，直接呼叫该 Skill 的名称或使用 /ai-twitter-digest 触发
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

v1.0.1

- Updated WhatsApp and Signal delivery target examples in documentation to use generic sample phone numbers. - No functional or code logic changes; documentation only.

v1.0.0

AI Twitter Digest automatically fetches tweets from a curated list of top AI researchers and tech leaders (Karpathy, Sam Altman, Jensen Huang, etc.), summarizes │ the day's key insights in Chinese using your preferred LLM, and delivers a clean two-part digest to your chat channel of choice.Supports Discord, WhatsApp, │ Telegram, Slack, and Signal out of the box. Includes an interactive setup wizard that auto-detects your local environment and writes the config for you — no │ manual editing required.

元数据

Slug ai-twitter-digest

版本 1.0.1

许可证 —

累计安装 0

当前安装数 0

历史版本数 2

常见问题