Skills

Wlzh Youtube Tracker

Track YouTube channels for new uploads.

Install

npx skills add wlzh/skills --skill wlzh-youtube-tracker -g -a claude-code -y
Repository
wlzh/skills694
Category
Engineering
Updated
8d ago
License
MIT
Context cost
Lightabout 1,244 tokens
Runs scripts
Yes, review before use

What it does

Track YouTube channels for new uploads. Supports both RSS mode (no API key needed, unlimited quota) and API mode. Use when: add/remove/list tracked YouTube channels, check for new videos, or run scheduled YouTube channel monitoring.

wlzh-youtube-tracker

A script-backed Skill to maintain a list of tracked YouTube channels and periodically check for new uploads.

v3.1.0 - Self-contained Telegram Sending (2026-06-06)

Critical reliability fix, script now sends Telegram directly, no LLM relay needed.

Changes

  • notify sends Telegram directly: Script calls Telegram Bot API itself, no LLM relay
  • Atomic seen marking: Only marks video as seen AFTER successful Telegram send
  • Auto-retry on failure: Failed sends stay "unseen", next cron cycle retries automatically
  • Reads bot token from ~/.openclaw/openclaw.json: No extra config needed
  • Fallback mode: If bot token unavailable, falls back to stdout output for LLM relay

Bug fix

  • v3.0.0 had a critical bug: notify marked videos as seen immediately upon detection, but Telegram sending was done by LLM. If LLM failed (rate limit/timeout), videos were lost forever, marked as seen but never sent. v3.1.0 fixes this completely.

v3.0.0 - Parallel Fetching + Notify Mode (2026-06-06)

Major performance optimization, parallel fetching, ~10x faster.

Changes

  • Parallel HTTP fetching: All channels now fetch concurrently (15 parallel curl processes)
  • ~10x faster: 55 channels check in ~4-5 seconds (was ~45 seconds with serial curl)
  • New notify command: Outputs Telegram-ready formatted messages (πŸ“ΊπŸŽ¬πŸ”— format)

  • Eliminates timeout failures (the #2 failure cause)
  • No more ~24k tokens consumed per check
  • Success rate expected: ~99% (was ~64%)

Why this matters

  • Previous cron used isolated agentTurn: LLM β†’ exec script β†’ LLM interpret β†’ send
  • Each run consumed 24-95k tokens and took 45-180s, frequently hitting API rate limits
  • New approach: systemEvent β†’ exec script (4s) β†’ send if needed β†’ done

v2.1.0 - Proxy Support (2026-06-05)

Read the whole SKILL.md on GitHub

More in Engineering