sota-scan
Is your project as good as the best out there? Find out.
Point it at your project. It asks one thing:
"What do the best projects have that I don't?"
It checks the top real projects in your space. Compares them to yours. Hands you a scorecard and a to-do list.
Every gap points to a real project that already does it. No vague advice. No fluff.
Why it's good
🪞 An honest mirror. Where you really stand.
🎯 A fair fight. It sorts projects by approach. It compares you to your real peers. Not a random pile. Borrowed ideas from other styles are marked optional, never counted against you.
🧾 Receipts, not opinions. Every gap names a real project.
✅ A real to-do list. Ranked. Worst first. With a next step.
📈 Tracks progress. Run it again. See what improved.
What you get
🧭 SOTA Standing — my-project
Verdict: Solid core. Behind on testing and security.
Score: █████░░░░░ 63% Standing: BEHIND vs 11 projects
▶ Do this next:
Add injection detection. Everyone serious has it. You don't.
Copy it from <real project>.
🏆 The leaders 📊 How you compare 🔧 To-do list
Try it (2 min)
It's an add-on for Claude Code. Copy two files in:
Mac / Linux
mkdir -p ~/.claude/skills/sota-scan ~/.claude/workflows
cp SKILL.md ~/.claude/skills/sota-scan/
cp workflows/sota-scan-fanout.js ~/.claude/workflows/
Windows
New-Item -ItemType Directory -Force ~/.claude/skills/sota-scan, ~/.claude/workflows | Out-Null
Copy-Item SKILL.md ~/.claude/skills/sota-scan/
Copy-Item workflows/sota-scan-fanout.js ~/.claude/workflows/
Open Claude Code in any project. Say:
/sota-scan
Or just ask:
- "Is my project top-tier?"
- "What am I missing vs the best?"
Done. It does the rest.
How deep?
| Say | You get | For |
|---|---|---|
/sota-scan quick |
top 3 gaps | a fast check |
/sota-scan |
full scorecard | everyday use |
/sota-scan exhaustive |
the deepest scan | launch / investor ready |
First time? Start with quick. Cheap, fast taste. Go deeper once you like it.
What it costs
It does real research — reads a dozen competitor projects, not just summaries. So it's not instant or free:
| Mode | Roughly |
|---|---|
quick |
~80–150k tokens · a minute or two |
standard |
~150–350k tokens · a few minutes |
exhaustive |
~400k+ tokens · several minutes |
Use it now and then — before a launch, a pitch, or when you're curious — not on every commit. On a Claude plan it's bundled. Paying per token? Lean on quick.
Share the result
Every scan is saved to .sota/ so you can re-run it and track progress. Want a copy you can drop in a PR, an issue, or a doc? Export a standalone Markdown report:
node scripts/sota-report.mjs --report # writes .sota/report.<domain>.md
You get a clean, self-contained file: summary + tier, the capability matrix, the high-confidence gaps, and a cited sources/disclosures trail. It's deterministic — the same scan always produces the same report — so --check flags a stale one, and no flag previews it to your terminal. See a real one: .sota/report.sota-benchmark-agent.md.
Good to know
- Needs Claude Code with internet.
- It won't tell you to pivot. Just how to win at what you already do.
- It scanned itself and came out top-tier — proof (and a shareable report) in
.sota/.
MIT licensed. Free to use and build on. See LICENSE.