# Competitor Research

> Runs verdict-based competitive research for a product or project idea before building it: candidate generation, parallel research agents, a strict empty/weak/strong-incumbent verdict rubric, honest kill decisions, and wedge sharpening for survivors. Use before starting any new project, adding a major feature, or when you suspect an idea is "surely already done" — it usually is, and the research tells you where the exception hides.

- Skill: `zhengqiuyang/competitor-research` (Agent Skill)
- Install (CLI): `npx skillmds@latest add zhengqiuyang/competitor-research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/zhengqiuyang/competitor-research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- License: MIT
- Author: zhengqiuyang (https://skillmd.com/u/zhengqiuyang)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/zhengqiuyang/competitor-research

---


# Competitor Research, With Verdicts

Most "competitive analysis" produces lists. Decisions need verdicts. This
method produces one of three per candidate, forces honest kills, and sharpens
only the survivors.

## Process

1. **Generate 3–6 candidates from accumulated signals** — not from a blank
   mind: gaps you noticed while building adjacent things, "why does nobody do
   X" moments, incumbents' sunsetting announcements, community complaint
   threads. Prior signals beat fresh brainstorming.
2. **Fan out one research agent per candidate**, each instructed to:
   - search aggressively with multiple query formulations (English + the
     market's other languages), GitHub API and npm/PyPI registry **facts**
     (stars, last-push dates, download counts) over marketing claims;
   - enumerate DIRECT competitors, adjacent tools, and **substitutes**
     (vendor built-ins are the stealth killer);
   - end with a verdict and a named strongest incumbent.
3. **Apply the rubric** (below). The verdict is about the SPACE, not your ego.
4. **Kill honestly.** A crowded space is a crowded space. Record the kill and
   the reason — it is cheap insurance against re-litigating the idea later.
5. **Sharpen the wedge for survivors only**: what exact cell of the matrix is
   empty? Write the one-sentence differentiation, then the honest kill
   criteria — which incumbent move, on what clock, erases you?

## The verdict rubric

- **(a) Empty** — nobody does this. Verify twice; empty cells are usually
  observation errors or demand illusions (an identical implementation sitting
  at 4 stars after 319 commits is a demand signal, not whitespace).
- **(b) Weak entries** — exists, but every player has a disqualifier:
  dormant (6+ months), single-platform, no maintenance, wrong shape. Winnable
  if you can name WHY they all stalled.
- **(c) Strong incumbent** — a well-executed tool with distribution owns it.
  Do not enter head-on. Only a documented, structural gap survives (e.g.
  "no lockfile-first flow" against a registry giant), and it carries a clock.

## Disciplines that keep it honest

- **Facts over vibes**: star counts and last-push dates from APIs; weekly
  downloads from registries. "Popular" is not a number.
- **Check the vendor's own roadmap surface**: the most common killer is the
  platform shipping your feature natively. Ask: why hasn't the vendor done
  this? (Answer sometimes: their business model points elsewhere — that gap
  is durable.)
- **Freshness warning**: a "underserved" conclusion older than ~2 months is
  suspect in fast markets. One observed case: a niche flagged empty was
  filled by an 11k-star incumbent's feature ship within eight weeks.
- **The substitute test**: for each incumbent ask "what do people use when
  they don't use this?" — the answer is often the real competitor.
- **Kill-tally sanity**: across 5 rounds and ~20 candidates, roughly
  three-quarters should die. If everything survives your research, your
  research is flattering you.