Sourcevine agent skills
Agent skills for public social data
Three ready-made workflows for an AI assistant: report on campaign posts, plan content tests from competitors’ posts, and find the questions in your comments.
Free to install. Sourcevine API usage consumes credits.
Codex + Claude CodeOne MCP connectionReadable skill files
Three skills
Start with your question.
Each skill includes a decision brief, an action plan and a report you can share.
“How are these campaign posts performing?”
Get a decision brief, measured changes and prioritized follow-ups from your campaign post links.
TikTokInstagramYouTubeX
- Bring
- Campaign post URLs · Optional previous snapshot
- Get
- HTML + Markdown report + JSON snapshot
10 posts · 20-credit estimated limit
“What are competitors posting, and what should we test?”
Turn a profile watchlist into original creative briefs, prioritized experiments and a clear learning plan.
TikTokInstagram
- Bring
- Profile URLs · Business context · Optional previous snapshot
- Get
- HTML + Markdown brief + CSV + JSON
5 profiles · 1 page each · 20-credit estimated limit
“What questions and concerns appear in these comments?”
Turn sampled comments into prioritized FAQs, draft answers and content ideas, with evidence and facts to confirm.
TikTokInstagramYouTube
- Bring
- Post URLs · Research question or business context
- Get
- HTML + Markdown brief + JSON evidence
3 posts · 1 page each · 24-credit estimated limit
Setup
Set up once in Codex or Claude Code.
- 01
Connect Sourcevine
Add your API key through MCP. All three skills use the same connection.
- 02
Install what you need
Save each skill file in its own folder. Choose one or keep all three available.
- 03
Ask a specific question
Add your URLs, context and budget. Keep the report and evidence in your workspace.
Follow the setup guide → Try before your first live request
Example reports.
Every skill has a fictional example you can inspect or analyze offline. No Sourcevine API credits are needed for those examples; your assistant’s own usage charges may apply.
Each run has a credit limit.
Each workflow estimates current API costs and limits its calls. Missing data stays visible, and a partial report explains what could not be checked.
Each finding links to its evidence.
Reports connect recommendations to sources, mark facts to confirm and include a shareable HTML brief. Public metrics do not establish conversions, and sampled comments do not represent an entire audience.