Imported from topviewai/skill (topview-youtube-kol-ops/SKILL.md). Install upstream with npx skills add topviewai/skill --skill topview-youtube-kol-ops. Copyright stays with the author.

Topview YouTube KOL Ops

Merchant-facing YouTube creator research via the host's unified topview-mcp server (progressive disclosure; domain youtube_kol).

Step 0 — Bootstrap

1. Confirm MCP server topview-mcp is connected (look for use_youtube_kol, get_tool_schema, call_tool).
2. If missing, point the user to mcp.example.json → ~/.cursor/mcp.json, reload MCP.
3. Do not run scripts/mcp_call.py, open SSE yourself, or use curl/raw HTTP.

Execution

Server: topview-mcp (unified). Never open SSE yourself.

YouTube KOL tools are deferred. Invoke via progressive disclosure:

1. Read the use_youtube_kol description on tools/list (embedded Tools catalog). Calling use_youtube_kol only to list tools is optional.
2. Pick a deferred youtube_kol_* name from the catalog or references/tool_routing.md.
3. get_tool_schema with { "name": "<youtube_kol_*>", "domain": "youtube_kol" }.
4. call_tool with { "name": "<youtube_kol_*>", "domain": "youtube_kol", "arguments": { "req": { ... } } } (confirm shape with schema).

Typical host tool ids: mcp__topview-mcp__get_tool_schema / mcp__topview-mcp__call_tool.
Prefer youtube_kol_auto_quote_saved over write tool youtube_kol_auto_quote unless the user explicitly wants a new saved quote.

UX Rules

1. Prefer analysis and shortlists over CRM execution.
2. Never invent subscriber/metrics; report empty/failed tool results honestly.
3. Creative production → topview-generate.
4. Match the user's language.

Source

Adapted from Marketing Studio youtube-kol-ops-assistant (V049).