diffmode-growth-tactics

{
  "name": "diffmode-growth-tactics",
  "description": "Diffmode's free growth-ideation pipeline as a Claude Code plugin. A /start orchestrator (point it at your product's URL, or run it bare and answer one question; re-running auto-resumes an unfinished run) drives the full DAG — fast diagnostics intake (URL research or minimal Q&A) -> 2-wave enrichment -> parallel think-tank research (competitor-gaps, cross-industry, platform-arbitrage), with per-run growth-factors mining (a clean-room LIGHT vector DB built fresh from public case studies) running concurrently from right after the competitors gate -> lite synthesis constraints -> a 2-stage synthesis: explore (blind vector combinations + emergent mechanisms) -> build (white-space ideation + founder-fit merge) -> a final synthesis.md of 7-9 novel demand-gen tactic IDEAS. It STOPS at synthesis: prioritization, implementation guides, and the proprietary 576-vector database stay paid. A parameterized reviewer gates two stages (enrichment competitors + the final synthesis build), score >= 7, max 3 retries; other generating stages get a structural check. Research stages run on the built-in WebSearch (zero setup, free, no API key), with every cited URL auto-verified for reachability; a Perplexity MCP is an opt-in alternative, routed to a separate worker only when the founder names it in the command. Analysis + synthesis stages run with no MCP by design. The run ends with a styled HTML report (WS/report/index.html, stdlib-rendered, opens in the browser) instead of bare markdown files. After the synthesis gate, a best-effort packaging stage re-writes the tactics as a founder-clean report (growth-tactics.md — plain tactic cards: what it is, why rivals can't copy it, how to start, when to kill it; no scores, no pipeline plumbing); the full synthesis ships alongside as a working paper, and the report falls back to it when packaging is skipped.",
  "version": "2.8.0",
  "author": {
    "name": "Anton Kogut"
  }
}