From research-tech-stack
Research a startup's real tech stack by triangulating GitHub repos, job postings, engineering blogs, and other sources. Use when asked about a company's technology, backend language, infrastructure, or engineering stack.
How this skill is triggered — by the user, by Claude, or both
Slash command
/research-tech-stack:research-tech-stackThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are researching the **real** tech stack of: **$ARGUMENTS**
You are researching the real tech stack of: $ARGUMENTS
Your goal is to determine what technologies this company actually uses in production — not what they market, not what their SDKs are written in, not what their frontend uses. The real backend, infrastructure, databases, and deployment stack.
Source: https://github.com/irinanazarova/skill-tech-stack-research
Before researching, do these first:
tech-stacks/{company-slug}.md in the current project directory. If it already exists, tell the user and ask whether to re-research or skip. Don't redo work silently.tech-stacks/ is in .gitignore. If not, add it — research output shouldn't be committed accidentally.WebSearch — searching the web for job postings, blogs, case studiesWebFetch — fetching job boards (Greenhouse, Ashby), blog posts, documentationBash(gh *) — GitHub CLI to list repos, check forks, read dependency filesBash(mkdir *) — creating the tech-stacks output directoryWrite — saving the final research fileTask — launching parallel research agentsBefore you begin, internalize these hard-won lessons from researching 47+ companies:
Companies hide their boring-but-real backend. Surge AI (Rails on Heroku), StackBlitz (Rails + PostgreSQL), Huntress (Rails primary), Figma (Ruby Sinatra), CrowdStrike (Rails for Risk Platform) — NONE of these advertise Rails. You must dig for it. Always ask: "What framework handles auth, billing, CRUD, and background jobs?"
87% of startups use TypeScript, but only 36% use it on the backend. React/Next.js is frontend — it tells you NOTHING about the server. When you see TS, verify: is it frontend-only, or are there actual Node.js backend services?
A Python SDK means their CUSTOMERS use Python. The platform itself could be Rails, Go, or anything else. Same for JS client libraries.
68% of companies "use Python" but only 17-23% use it as their web backend (Django/FastAPI/Flask). The rest use it for ML training, scripting, and automation. Distinguish between "Python for ML pipelines" and "Python as the web backend."
A drone company does NOT mean Python. A fintech does NOT mean Java. An AI company does NOT mean Python backend. Industry category tells you NOTHING about the web backend framework. dScribe AI builds autonomous drones with computer vision — their backend is Ruby on Rails. Figma is a design tool — their backend is Ruby Sinatra. Domain-based guesses are wrong more often than they're right. If you have no evidence, say "Unknown" — do NOT guess based on what seems typical for the industry.
A React/CRA/Next.js frontend with no Turbolinks, no CSRF meta tags, and no Rails asset fingerprinting does NOT mean "not Rails." Many Rails apps serve as API-only backends behind decoupled React SPAs. In this architecture, you will see ZERO Rails indicators in the HTML. Finta (YC) has a React CRA + Ant Design frontend — but is confirmed Rails via sfruby.com job posting and SF Ruby conference demo. NEVER conclude "not Rails" based solely on frontend evidence.
Forked repos reveal real infrastructure: sidekiq-cron = Ruby backend, celery forks = Python, devise/clearance = Rails auth, ember-toucan = Ember.js. Check forks BEFORE original repos. See signals.md for the full fork cheat sheet.
Greenhouse and Ashby have the most technical detail. Generic careers pages and LinkedIn have the least. Check ALL ATS platforms, not just the company website.
Execute these phases IN ORDER. Do as many searches in parallel as possible within each phase.
Search for the company's GitHub org. For each org found:
Search these sources simultaneously:
"{company}" software engineer jobs (web search — add the current year to the query)job-boards.greenhouse.io/{company} (direct ATS)jobs.ashbyhq.com/{company} (direct ATS)site:builtin.com "{company}" engineer (web search)"{company}" OR "{domain}" backend engineer requirements (web search)site:news.ycombinator.com "{company}" for HN Who's Hiring postssite:sfruby.com "{company}" — SF Ruby community job board (high signal for hidden Rails companies)site:rubyonremote.com "{company}" — Ruby-specific remote job boardFor EACH job posting found, extract:
Search in parallel:
"{company}" site:rubyevents.org (Ruby conference/meetup talks)"{company}" site:railsconf.org OR site:rubyconf.org (major Ruby conferences)"{company}" site:gophercon.com OR site:golangevents.com (Go conferences)"{company}" site:rustconf.com (Rust conferences)"{company}" founder OR CTO meetup talk (web search)Conference talks are extremely high signal — founders presenting at language-specific meetups confirms the stack. dScribe AI (a drone company) was found to be Rails because their CTO presented at SF Ruby meetup.
Search in parallel:
"{company}" engineering blog tech stack (web search)site:{domain} engineering OR blog architecture OR infrastructure (web search)"{company}" CTO podcast tech stack (web search)"{company}" site:newsletter.pragmaticengineer.com (web search)"{company}" site:blog.bytebytego.com (web search)Look for:
Search in parallel:
"{company}" site:cloud.google.com (case studies)"{company}" site:aws.amazon.com (case studies)"{company}" software engineer ruby OR rails OR python OR golang linkedin (engineer profiles)"{company}" tech stack site:himalayas.app (aggregator — weak signal)deepwiki.com/{github-org}/{repo} if open-sourceSearch: "{company}" OR "{domain}" heroku OR sidekiq OR celery OR kafka OR kubernetes
See signals.md for infrastructure → backend mapping and category calibration tables.
After gathering evidence, apply these checks:
See template.md for the full output template.
Save results to tech-stacks/{company-slug}.md (create the directory if it doesn't exist). The slug should be lowercase, hyphenated (e.g., crowdstrike.md, surge-ai.md).
tech-stacks/{company-slug}.md.npx claudepluginhub irinanazarova/skill-tech-stack-research --plugin research-tech-stackDiscovers a target's technology stack via passive OSINT signals. Routes to sub-skills for frontend, backend, infra, security, and correlation analysis.
Recommends full technology stacks for new projects based on requirements, team expertise, and constraints like scale, performance, budget, and time-to-market.
Opinionated tech stack recommendations and settled library references. Runs parallel researcher agents to compare candidates and emit a stack report + ADRs for greenfield projects, or routes direct references for coding with settled defaults.