MCP Opportunity Pipeline Skill

Automatically activate when the user asks about:

• Finding MCP or automation tool opportunities
• Validating product ideas against Reddit/forums
• Building MCP servers for Apify or similar platforms
• Analyzing marketplace gaps
• Publishing to Apify, Smithery, or other MCP marketplaces

When to Invoke

Discovery & Research

Triggers: "find opportunities", "market research", "what should I build", "MCP ideas" Action: Use /mcp-pipeline:discover to scrape marketplaces

Gap Analysis

Triggers: "analyze gaps", "score opportunities", "which is best" Action: Use /mcp-pipeline:analyze-gaps on discovery results

Validation

Triggers: "validate idea", "check Reddit", "real demand", "people want" Action: Use /mcp-pipeline:validate against Reddit pain signals

Specification

Triggers: "write spec", "plan build", "what features" Action: Use /mcp-pipeline:spec for validated opportunities

Building

Triggers: "build it", "implement", "create actor", "scaffold" Action: Use /mcp-pipeline:build from spec

Testing

Triggers: "test it", "QA", "does it work", "verify" Action: Use /mcp-pipeline:qa on built code

Publishing

Triggers: "publish", "deploy", "ship it", "launch" Action: Use /mcp-pipeline:package then /mcp-pipeline:publish

Full Pipeline

Triggers: "full pipeline", "end-to-end", "run everything" Action: Use /mcp-pipeline:run with appropriate parameters

Pre-Check Behavior

Before running expensive operations:

1. Check for existing outputs:

   outputs/discover/    → Skip discover if recent data exists
   outputs/analyze/     → Skip analyze if recent analysis exists
   outputs/validate/    → Skip validate if recent validation exists
   

2. Suggest incremental runs:

   • If discovery exists but not analysis, suggest /mcp-pipeline:analyze-gaps
   • If validation exists but not spec, suggest /mcp-pipeline:spec

3. Warn about costs:

   • Discovery scrapes multiple sites (rate limits)
   • Validation searches Reddit (API usage)
   • Building creates deployable code

Configuration Reference

Load from .claude-plugin/config.json:

• Marketplace settings
• Subreddit lists
• Scoring weights
• QA thresholds
• Publishing settings

Output Locations

| Stage | Output Path | |-------|------------| | Discover | outputs/discover/raw-opportunities-{date}.json | | Analyze | outputs/analyze/gap-opportunities-{date}.json | | Validate | outputs/validate/validated-opportunities-{date}.json | | Spec | outputs/spec/{name}-spec.md | | Build | outputs/build/{name}/ | | QA | outputs/qa/{name}-qa-report.json | | Package | outputs/package/{name}/ | | Publish | outputs/publish/publication-log.json |

Example Conversations

"I want to find MCP opportunities"

→ Check if recent discovery exists → If not, run /mcp-pipeline:discover --phase casual → Then suggest /mcp-pipeline:analyze-gaps

"Which opportunity should I build?"

→ Check if analysis exists → Present top opportunities with scores → Suggest validation for top picks

"Build the Notion sync tool"

→ Check if spec exists → If not, generate spec first → Run /mcp-pipeline:build --name notion-sync → Then run QA

"I want to ship something to Apify this week"

→ Run full pipeline: /mcp-pipeline:run --phase casual --target apify → Stop at each stage for approval → End with publish (dry-run first)

Error Recovery

| Error | Recovery | |-------|----------| | No discover data | Run discover first | | Low validation scores | Pivot to different opportunity | | QA failures | Fix issues, re-run build + QA | | Publish failure | Check credentials, retry |

Important Notes

• Always check existing outputs before re-running stages
• Validate ideas before building (save time)
• Use dry-run before actual publish
• Marketing posts require manual review/submission