메뉴
Run the YouTube insight refresh pipeline — fetch transcripts from tracked channels, extract insights, and deduplicate. Use when scheduled or when asked to refresh insights.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Generate presentation slide decks from markdown content using the Gamma API. Use when user says "create slides", "make a presentation", "build a deck", "generate a slide deck", or asks to turn content into slides. Run with /gamma-slides.
Package any combination of skills into a distributable Claude Code / Cowork plugin. Builds the correct directory structure, rewrites frontmatter, creates slash commands, generates marketplace.json, and zips for upload. Trigger with "build a plugin", "package these skills as a plugin", "create a plugin from", or "turn these skills into a plugin".
Create or update a skill that packages a workflow into a self-contained SKILL.md with scripts, references, and assets. Use when user says "create a skill", "make a new skill", "crystallize this workflow", "turn this into a skill", or asks to build a reusable process.
Debug container agent issues. Use when things aren't working, container fails, authentication problems, or to understand how the container system works. Covers logs, environment variables, mounts, and common issues.
Generate images using Gemini via scripts/generate-image.js. Creates technical diagrams, architecture visuals, blog headers, and branded graphics. Use when the user asks to "create an image", "generate a diagram", "make a visual", or "create a graphic". Triggers on "generate image", "create image", "make diagram", "create visual", "architecture diagram".
Generate images using Gemini via scripts/generate-image.js. Creates technical diagrams, architecture visuals, blog headers, and branded graphics. Use when the user asks to "create an image", "generate a diagram", "make a visual", or "create a graphic". Triggers on "generate image", "create image", "make diagram", "create visual", "architecture diagram".
| name | refresh-insights |
| description | Run the YouTube insight refresh pipeline — fetch transcripts from tracked channels, extract insights, and deduplicate. Use when scheduled or when asked to refresh insights. |
| allowed-tools | Bash(*), Read, Write, Edit, Glob, Grep, mcp__bastionclaw__* |
Run the full insight extraction pipeline: fetch transcripts → extract insights → deduplicate.
Read the channel list and lookback period:
cat /workspace/project/.claude/skills/youtube-planner/sources.json
This file contains:
sources: array of YouTube @handles to fetchlookback_days: how many days of videos to pullIf the file doesn't exist or is empty, report the error via send_message and stop.
For each channel in sources, fetch recent videos:
python3 /workspace/project/.claude/skills/youtube-planner/catch-up-channel.py @CHANNEL LOOKBACK_DAYS
TRANSCRIPT_API_KEY must be in the environment (passed via container secrets).
After all channels are fetched, update lookback_days to 1 so future runs only pull the last day:
# Copy to writable location, modify, copy back
cp /workspace/project/.claude/skills/youtube-planner/sources.json /workspace/group/sources.json
# Edit lookback_days to 1 in /workspace/group/sources.json
cp /workspace/group/sources.json /workspace/project/.claude/skills/youtube-planner/sources.json
If the write back fails (read-only filesystem), skip it — non-critical.
Find all new transcript.json files not yet indexed in insight_sources. For each one:
add_insight MCP tool. Pass source_metadata as a JSON string with author, published, viewCount, videoId.search_insights or link_insight_source.Call the dedup_insights MCP tool and WAIT for it to return. The response contains an output field with the exact merge count and remaining insight count. Parse these numbers — do not paraphrase or guess.
Send a summary via send_message that MUST include ALL of these:
dedup_insights response