Follow Builders: AI Builders Digest
Tracks the top builders in AI — researchers, founders, PMs, and engineers — and delivers curated daily or weekly summaries of their X posts and podcast appearances, in English, Chinese, or bilingual.
The skill clearly follows least privilege: no user API keys needed, all content fetched centrally, Telegram/Email keys stored locally in ~/.follow-builders/.env and only if chosen. Data flow is transparent, explicitly states no API keys are sent, only public content read. Config stored locally, rollback possible (delete prompts files to reset). No malicious behavior found, but publisher is unverified, relies on third-party central service (potentially unaudited code), and permission boundaries depend on user confirmation/selection, hence deduction.
Instructions are internally consistent, with clear steps and error handling guidance, but the scripts (prepare-digest.js, deliver.js) are not included in reviewed files, so their availability cannot be verified. Depends on external central feed network availability, and no tests or debug data provided. Thus, happy path plausible but reliability evidence thin, deduct.
Target audience and scenarios clear (AI builder digest, works with OpenClaw or non-persistent agents), detailed triggers (/ai) and environment adaptation (OpenClaw detection, crontab setup). Boundaries clear (no source list modification, no web scraping). Chinese support explicit (language setting includes Chinese or bilingual). However, core function depends on central feed which may be unreachable from mainland China (no domestic alternative mentioned), hence deduction.
Documentation is well-structured, with README (EN/ZH), clear installation steps, configuration guide, customization instructions. But lacks explicit version number, changelog, FAQ, and known-limitations section. License is MIT, but publisher unverified, maintenance responsibility and update path unclear. Therefore, information architecture good but governance incomplete, deduct.
Core concept (digest generation) clear, output format has examples, but no actual sample output (examples/sample-digest.md not provided), and depends on real-time central feed; static review cannot verify output correctness/completeness. Value proposition clear, but direct usability evidence weak; per static review cap of 7, give 7.
Key operations (fetch, remix, deliver) depend on external scripts and central feed; not independently verifiable from source. README claims sample output but not provided. No test suites or CI workflows. Per static review cap of 5, and actual evidence thin, give 5.
- Core function depends on central feed, which may not be accessible from mainland China; verify reachability or provide domestic alternative.
- Central feed's quality, security, and data authenticity are unverified; reliance on third-party service poses risks.
- Publisher is unverified; trust unknown, recommend reviewing repository history and code before installation.
- Script files not included; actual behavior of prepare-digest.js and deliver.js cannot be confirmed; review before use.
What does this skill do, and when should you use it?
Follow Builders is a content curation skill that aggregates the latest from a curated list of AI builders — researchers, founders, product managers, and engineers — across their X/Twitter posts and multiple AI podcast transcripts. All content is fetched centrally and served via a public feed, so no API keys are required from the user. The skill remixes the raw content into digestible summaries, with output in English, Chinese, or bilingual (paragraph-by-paragraph). It delivers via OpenClaw's channel system, Telegram, or email, or displays directly in-chat when supported.
The skill runs Node.js scripts to fetch a central feed containing podcast transcripts and X posts, then generates a configuration based on user preferences (frequency, time, timezone) and delivery method (Telegram, email, or stdout). It walks the user through an onboarding flow, sets up cron jobs (using openclaw cron add for OpenClaw, or system crontab for other environments), and remixes tweets and podcast transcripts into clean summaries with links to the original content. All fetching and summarization uses JSON data from the feed; no web searches or API calls are made.
- AI product managers who want a concise weekly digest in Chinese of what top AI founders are saying on X, delivered every Monday morning.
- AI researchers who want to stay on top of new research and coding tips shared by researchers, engineers, and founders without scrolling through many posts.
- OpenClaw users who want a daily digest automatically delivered to their Telegram or Discord channel without any extra setup.
- Users in a non-persistent environment (e.g., Claude Code) who prefer an on-demand digest via the `/ai` command instead of a scheduled job.
- Non-native English speakers who want bilingual output to preserve the original English context while reading summaries in Chinese.
- Users who want to customize the tone and length of summaries, either through conversation or by editing the prompt files.
What are this skill's strengths and limitations?
- No API keys needed for content — all fetching is done centrally
- Three language modes: English, Chinese, and bilingual with paragraph-by-paragraph interleaving
- Settings are easily adjusted via conversation without editing config files
- Curated source list including 26 X builders, 6 podcasts, and 2 official blogs
- Works on OpenClaw and other non-persistent agents (via Telegram/email delivery)
- Clear onboarding flow with step-by-step guidance for non-technical users
- Requires Node.js and running `npm install`
- Relies on availability of the central feed — if it's down, the skill fails
- For non-OpenClaw environments, cron delivery bypasses LLM remixing, sending raw JSON instead of a polished digest
- Customizing summary prompts requires manual file editing or relying on the agent to correctly interpret settings changes
- Telegram/email delivery requires the user to set up a bot or Resend account, adding friction
- No automated tests and no explicit testing on non-OpenClaw platforms
How do you install this skill?
Clone the repository into your skills directory:
# OpenClaw
git clone https://github.com/zarazhangrui/follow-builders.git ~/skills/follow-builders
cd ~/skills/follow-builders/scripts && npm install
# Claude Code
git clone https://github.com/zarazhangrui/follow-builders.git ~/.claude/skills/follow-builders
cd ~/.claude/skills/follow-builders/scripts && npm installCLI installation (clawhub install follow-builders) is coming soon.
How do you use this skill?
After installation, trigger the skill in your agent:
- Say "set up follow builders" or type /follow-builders to start the onboarding.
- The onboarding will ask for frequency (daily or weekly), time, timezone, language (English, Chinese, or bilingual), and delivery method (Telegram, email, or on-demand only).
- For delivery, follow the precise steps in the onboarding (e.g., creating a Telegram bot via BotFather and getting the chat ID).
- After setup, your first digest is sent immediately, and subsequent ones arrive on schedule.
- Use /ai anytime to manually trigger a digest.