菜单
Extract transcripts from any video URL — YouTube, TikTok, Instagram, X, Facebook, or direct file links. Also fetches metadata (title, author, stats). Supports batch transcripts, translation, and YouTube search.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
The 1-Prompt Meta Ads Engine. End-to-end Meta ad generation + paused campaign creation workflow. Asks 4 discovery questions, then runs the full pipeline: competitor intel (claude-ads /ads competitor), brand-site scraping for context, bulk image generation (kie.ai gpt-image-2), ad copy (claude-ads /ads creative), and paused Meta campaign creation (Meta Ads MCP). Dynamic Product Ad principles (brand compliance + variant matrix) are applied inline. Saves brand profiles for one-command re-runs. Trigger phrases (auto-fire on any of these in user prompt): - "/ad-engine" or "/ad engine" - "run the ad engine" - "launch ads for [brand]" / "build ads for [brand]" / "generate ads for [brand]" - "build a Meta campaign for [brand]" / "build the campaign for [brand]" - "20 ads for [brand]" / "20 Meta ads for [brand]" - "automate Meta ads" / "automate my Meta ads" - "one prompt ads" / "one-prompt ads" - "fresh ad batch" / "generate a fresh ad batch"
Use when creating Instagram carousels from a reel script, handout, PDF, article, or rough idea. Triggers on "carousel", "Instagram carousel", "IG carousel", "turn this script into slides", "build a carousel", "make slides for IG", "anchor slide", "visual brief". Runs the reference-first workflow (copy → brief → 3 anchor variants → locked anchor → matched slides one at a time → consistency check) and generates real images via the kie.ai gpt-image-2 MCP. Skip for plain captions or non-designed multi-slide posts.
Generate a hand raiser PDF lead magnet (1-5 pages) for any script or standalone topic. Renders a professional guide/cheatsheet/checklist, uploads to Google Drive, and links in Notion.
One-prompt digital footprint removal. Finds where a person's personal data (home address, phone, email, relatives) is exposed across data brokers and people-search sites, checks for breaches, ranks everything by risk, and auto-writes every CCPA/GDPR opt-out and deletion request filled and ready to send, then saves it all as a personal removal kit. Removes the manual searching and letter-writing. Triggers on: "/wipe-footprint", "wipe my footprint", "wipe my digital footprint", "erase me from the internet", "delete my data", "remove my data from data brokers", "stop the spam calls", "opt me out of data brokers".
Use when running inside Claude Design specifically. Produces a valid, brand-accurate HyperFrames video draft using pre-valid skeletons. For Claude Code / Cursor / Codex, use the `hyperframes` skill instead.
Hourly Instagram analytics pipeline that pulls per-post and account-level insights from the Meta Graph API into Supabase, mirrors live values into a Notion dashboard, and runs a daily Claude routine that classifies each post into Outlier/Average/Underperformer tiers with a 2-3 sentence "why" note. Built for content creators who want feedback that compounds, not a vanity follower counter. Triggers on "ig analytics", "instagram analytics", "performance dashboard", "content analytics pipeline", "instagram graph api", "outlier post detection", "tier classification", "creator analytics".
| name | transcript |
| description | Extract transcripts from any video URL — YouTube, TikTok, Instagram, X, Facebook, or direct file links. Also fetches metadata (title, author, stats). Supports batch transcripts, translation, and YouTube search. |
Extract transcripts from any video platform with one command. Paste a URL, get the transcript.
~/.config/supadata/.env (contains SUPADATA_API_KEY=...)https://api.supadata.ai/v1x-api-key: <key>Before making API calls, read the key:
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
Use this variable in all curl commands below.
Extract from user input or $ARGUMENTS:
text (plain text, default) or timestamped (chunks with timestamps)If no URL provided, ask:
What video do you want the transcript for? Paste the URL.
Supported: YouTube, TikTok, Instagram, X/Twitter, Facebook, or any direct video/audio file URL.
If only a URL is provided, default to transcript + metadata, plain text format. Don't over-ask.
Detect the platform from the URL to route correctly:
| URL Pattern | Platform |
|---|---|
youtube.com, youtu.be, youtube.com/shorts/, youtube.com/live/ | YouTube |
tiktok.com, vm.tiktok.com | TikTok |
instagram.com/reel/, instagram.com/p/, instagram.com/tv/ | |
twitter.com, x.com | X/Twitter |
facebook.com, m.facebook.com | |
.mp4, .webm, .mp3, .wav, .flac, .m4a, .ogg, .mpeg | Direct File |
youtube.com/playlist, playlist ID | YouTube Playlist |
youtube.com/@, youtube.com/channel/ | YouTube Channel |
| Everything else (non-video URL) | Web page → use /web/scrape fallback |
This is the primary flow — one URL, one transcript.
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
URL="THE_VIDEO_URL"
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/metadata?url=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$URL', safe=''))")"
Display a header with the metadata:
**{title}**
{author} · {platform} · {duration} · {views} views · {likes} likes
Published: {date}
For YouTube URLs, use the dedicated YouTube endpoint (more features):
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/youtube/transcript?url=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$URL', safe=''))")&text=true"
For all other platforms (TikTok, Instagram, X, Facebook, files), use the universal endpoint:
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/transcript?url=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$URL', safe=''))")&text=true"
If the API returns HTTP 202 with a jobId, the transcript is being generated asynchronously (common for non-YouTube platforms and long videos):
# Poll every 5 seconds until complete
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/transcript/$JOB_ID"
Poll until status is completed or failed. Max 12 attempts (60 seconds).
Show the full transcript in the conversation.
Save to file:
# Filename: ~/ai-content-system/output/transcripts/{platform}-{sanitized-title}-transcript.md
The saved file should include:
# {Title}
**{Author}** · {Platform} · {Duration} · {Views} views
**URL:** {original URL}
**Extracted:** {date}
---
{full transcript text}
If user asks for timestamps, use text=false in the API call. The response returns chunks:
[{"text": "chunk text", "offset": 0, "duration": 5000, "lang": "en"}]
Format as:
[0:00] chunk text
[0:05] next chunk
When user asks to translate a transcript:
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/youtube/transcript/translate?url=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$URL', safe=''))")&lang=$LANG_CODE&text=true"
Note: Translation takes 20+ seconds and costs 30 credits/minute. Warn the user about the higher cost if the video is long.
Common language codes: es (Spanish), fr (French), de (German), pt (Portuguese), ja (Japanese), ko (Korean), zh (Chinese), ar (Arabic), hi (Hindi), it (Italian).
For playlists, channels, or multiple videos:
source ~/.config/supadata/.env
# For a playlist:
curl -s -X POST -H "x-api-key: $SUPADATA_API_KEY" -H "Content-Type: application/json" \
-d '{"playlistId": "$PLAYLIST_URL", "text": true, "limit": 10}' \
"https://api.supadata.ai/v1/youtube/transcript/batch"
# For a channel:
curl -s -X POST -H "x-api-key: $SUPADATA_API_KEY" -H "Content-Type: application/json" \
-d '{"channelId": "$CHANNEL_URL", "text": true, "limit": 10}' \
"https://api.supadata.ai/v1/youtube/transcript/batch"
# For multiple video IDs:
curl -s -X POST -H "x-api-key: $SUPADATA_API_KEY" -H "Content-Type: application/json" \
-d '{"videoIds": ["id1", "id2", "id3"], "text": true}' \
"https://api.supadata.ai/v1/youtube/transcript/batch"
Then poll for results:
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/youtube/batch/$JOB_ID"
Save each transcript as a separate file in ~/ai-content-system/output/transcripts/batch-{topic}/.
When user wants to find and transcribe videos on a topic:
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/youtube/search?query=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$QUERY', safe=''))")"
Show the results and ask which video(s) to transcribe. Then run the single transcript flow for each selected video.
If the URL is not a video platform, fall back to web scraping:
SUPADATA_API_KEY=$(grep SUPADATA_API_KEY ~/.config/supadata/.env | cut -d= -f2)
curl -s -H "x-api-key: $SUPADATA_API_KEY" \
"https://api.supadata.ai/v1/web/scrape?url=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$URL', safe=''))")"
Returns the page content as markdown. Useful for blog posts, articles, docs.
| HTTP Code | Meaning | Action |
|---|---|---|
| 200 | Success | Display transcript |
| 202 | Async job started | Poll with jobId until complete |
| 206 | No transcript available | Tell user: "No captions available for this video. The video may be too short, have no speech, or captions are disabled." |
| 400 | Bad request | Check URL format |
| 401 | Invalid API key | Tell user to check ~/.config/supadata/.env |
| 402 | Out of credits | Tell user to check their SupaData plan |
| 404 | Video not found | Tell user: "Video not found — it may be private, deleted, or the URL is wrong." |
| 429 | Rate limited | Wait 10 seconds, retry once |
After showing a transcript, offer relevant next steps based on context:
If the user is working on content (e.g., came from /content-master), proactively suggest how the transcript could inform their scripts.