The antidote to “You’re absolutely right!”
Tired of your agent confidently saying the job is done, only to find it missed the mark again, and going around in circles correcting it? Spotter Round makes it prove the work before it tells you it’s finished.
$39. Every pack, now and later: $149.
Works with
Fit check
Not sure which pack fits? Find yours in two questions
Every requirement, in detailSound familiar?
Tap every one you’ve lived through.
Nothing yet? Lucky you.
Why it keeps happening
The agent that made the change is the one telling you it worked, from the same context and with the same blind spots. It isn’t lying. It checked part of the page and reported on the whole page.
So you become the check. You open the page on your phone, find what it missed, explain it again, and wait. Three rounds later you’re doing the job you handed it, and wondering what else it called done.
Drag the handle
The fix
Spotter Round is a small kit of skills your agent loads and scripts it runs before it tells you it’s done. You get a pass or a fail you can read, with a count of everything it checked, instead of a confident sentence.
Runnable scripts measure the layout, the pixels, the speed and the live site, and print what they looked at.
Every check ships a broken example it has to catch first, so a green result means something.
From the real build at real screen sizes to what the public actually gets after the deploy.
The same task, run twice. First on confidence. Then with one check fired first, and the aim corrected from where that shot landed.
A silent demonstration, shown as an agent conversation beside a paper shooting target on a night range; whichever side is acting is framed in brass, and the target is dimmed with a waiting note while the conversation runs. An agent is asked to fix a checkout bug and verify it. It answers “Fixed and verified. All checks passed,” but three rounds punch through the paper far outside the rings, circled and labeled “where it landed.” The user replies “You definitely missed,” and the target is stamped REPORTED HIT. The run restarts. This time the agent runs Skill(spotter-round:verification-protocol) and puts a single spotter round downrange first; it lands low and left of center and is circled as the spotter round. An arrow measures the difference back to the middle, labeled “correct from the difference.” The next four rounds land tightly grouped in the center of the target, which is stamped VERIFIED HIT. The agent reports that one check fired first and it corrected from where that shot landed, and the user replies “That’s perfect.”
Silent, 30 seconds, on a loop. Drag the rail to step through it, or slow it to half speed.
Who it’s for
The skills, grouped by the job they do. Start with the one that fits the change you just made.
Deployment checklist
Your agent says the layout passed. The sweep reads every line box and every
edge.
Deployment
Your agent says the layout passed. The sweep reads every line box and every edge.
FitsDeployment checklist
Your agent says the layout passed. The sweep reads every line box and every
edge.
Deployment
Your agent says the layout passed. The sweep reads every line box and every edge.
FitsA real build, at real screen sizes, in light and dark, with a clean browser console. No cut-off text, nothing spilling off the edge, and the design didn't move when nobody meant it to.
Builds your previous version beside the new one, serves both the same way, and measures both. The two numbers can honestly be compared, and it tells you when they can't.
/robots.txt Fetched twice
Verdict: stale. Purge the cache, don't redeploy.
An old copy and a deploy that never landed need opposite fixes, so each gets its own verdict.
After the deploy, it checks what the public actually gets. Every address is fetched twice, so an old copy a cache is still handing out can't pass for the new build, on any host.
Every skill, with its job in one line. Each one works on any framework and any host.
spotter-start
Which check fits the change you just made.
verification-protocol Also free on its own
A real build, real screen sizes, both light and dark, and a clean browser console.
layout-proof
No cut-off text, nothing spilling off the screen — and the check isn't fooling itself.
visual-fidelity-gate
Proves the design didn't move: same pixels, three browsers, nothing jumping as the page loads.
build-ab-harness
Faster than last time, measured so the two runs can honestly be compared.
post-deploy-proof
Every address fetched twice, so an old copy a cache is still handing out can't pass for the new build.
One download: skills for Claude Code, the same instructions to paste into any tool that reads AGENTS.md, and checks that prove the build, the layout and the live site are what you think they are.
A zip of plain folders. Each skill is a SKILL.md your agent reads, and most carry short, commented scripts it runs.
spotter-start is the router: it reads what you changed and names the one check that fits, so you never have to guess which skill to load.
A check that has never failed is a green light with extra steps. Each one here ships with a deliberately broken example it has to catch. Run that first: if the check calls the broken example fine, don't trust it on your real work either.
Run the broken examples
# run from the project you are checking, with Spotter Round unzipped at /path/to/spotter-roundnode /path/to/spotter-round/skills/layout-proof/scripts/orphans.mjs --selftest # plants a known problem and must catch itnode /path/to/spotter-round/skills/visual-fidelity-gate/scripts/diff.mjs --dir /path/to/spotter-round/examples/negative-control-diff # must go rednode /path/to/spotter-round/skills/post-deploy-proof/scripts/served.mjs --selftest # must tell a stale copy from a missing deployWe call these negative controls: a planted defect the check has to fail on. The first command builds a known stranded word and asserts the detector catches it; the second points the picture comparison at a pair that differs on purpose and must go red; the third serves a stale copy and a missing deploy and must tell them apart.
[ Gate ] visual-fidelity-gate //
# from the project you are auditingnode /[ Result ]fidelity: ALL OK//
[ Control ] negative-control-diff //
# from the project you are auditingnode /[ Result ]2 FAIL · exit 1//
[ Output ] excerpt //
FAIL card-320 SIZE MISMATCH 320×200 vs 320×256 — layout changedFAIL tint-320 mean |Δ| 15.40/255 > 12 pair(s) · worst mean 30.81/255 · threshold 12 FAIL# expected: 2 FAIL, exit 1. If this reports clean, the gate is not measuring anything.
[ Shot 01 ]
orphans.mjs --selftest
Aimed: plant a known problem; the check must catch it
ok detector: separates the twins: found 1 orphan, expect exactly 1
ok detector: flags the right one: "eeee", expect "eeee"
ok (1) the in-page control runs and passes in both passes
33 cases behaved as designed
Landed where it was aimed // exit 0
[ Shot 02 ]
diff.mjs --dir …/negative-control-diff
Aimed: planted differences; the check must go red
FAIL card-320 SIZE MISMATCH 320×200 vs 320×256 — layout changed (resized diff: mean 30.81)
FAIL tint-320 mean |Δ| 15.40/255 > 1 ≥16 off 29.43% ≥32 off 0.00%
2 pair(s) · worst mean 30.81/255 · threshold 1
2 FAIL
Landed where it was aimed // exit 1
[ Shot 03 ]
served.mjs --selftest
Aimed: a stale copy and a missing deploy; the check must tell them apart
ok a stale cached copy is caught (expected stale, got stale)
ok a behind deploy is caught (expected origin, got origin)
ok a stale cache and a behind deploy get different verdicts
28 cases behaved as designed
Landed where it was aimed // exit 0
The check that has never failed is the one to distrust.
The skills are plain text files your agent reads. The scripts are short, commented, and run only on your computer.
claude --plugin-dir /path/to/spotter-roundThe written skills need nothing at all. The scripts need Node 18 or newer, plus two packages in the project you're checking:
npm i -D playwright sharp && npx playwright install chromium webkit firefoxShort scripts with comments, and nothing compressed into a blob. A product that says “don't trust a report you can't read” has to be readable itself.
Spotter Round
$39.
Please don't resell it or share it outside your company.
All-Access Bundle
Spotter Round, the Agent Setup Pack, the Email Pack, the WordPress Pack, the App Stack Pack and the Campaign Tracker Pack, plus every pack added after them, each with every future version.
Spotter Round if an agent writes your web pages: it proves the work before and after it ships. The Agent Setup Pack if you use more than one agent or more than one tool, or hand agents API keys. The Email Pack if you send HTML email. The WordPress Pack if you build on WordPress with Bricks. The App Stack Pack if you build web apps on Cloudflare Pages, Supabase or Stripe, or use SureContact or OttoKit on any site, WordPress included. The Campaign Tracker Pack if you run ads and want short links, leads, purchases and cost per lead tracked from the server, on any site behind Cloudflare. The All-Access Bundle is every pack, now and later, for less than the six together.
Yes. The skills are plain text files and the checks are plain Node scripts, so any agent can run them. Claude Code loads the skills from its skills folder. Codex, Cursor and the rest take a skill pasted into your repository's AGENTS.md, and Gemini CLI reads that file once its settings name it.
No tracking, no license check, no analytics, nothing phoning home. The scripts only visit addresses you type, with one exception: the speed comparison runs Lighthouse through npx, which fetches it from the npm registry unless your project already has it. You can confirm that yourself: the scripts are short enough to read.
/verify. Why buy this?Different jobs, and they work together. /verify is a built-in, prompt-based skill that builds and runs your app “without falling back to tests or type checks”, and it runs only when you invoke it. Anthropic's own docs say that working out how to launch your app “gets unreliable” once a project needs a database, an env file, a graphical session or a multi-step build. This is runnable scripts instead of a prompt: the same checks every time, a pass or fail you can wire into CI, a broken example proving each check can fail, and things /verify doesn't mention at all — before-and-after pictures, three-browser checks, fair speed comparisons, and proof that the live site serves the build you made.
Into their own packs, so each one can do a single job well: the Agent Setup Pack, the Email Pack and the App Stack Pack. The skills kept their names, and each pack ships its own checks and the broken examples that prove them.
More than you paid for: everyone who bought Spotter Round before version 2 gets the All-Access Bundle, every pack now and every pack we add later.
Numbered downloads with a list of what changed. Every future version of Spotter Round is included at no charge, so you never pay for a new version, and you are never forced to update.
Yes — that's what it's for. One company covers every project that company delivers.
Fourteen days, no questions. A product about checking things that argued with you over a refund would be a poor advertisement for itself.
Sold separately
Each pack is sold on its own and ships its own checks, the same way this one does.
One set of rules, proven to reach every coding agent, several agents on one project without landing work on each other's branches, and API keys handed over without entering the chat.
HTML email that survives real inboxes, dark mode and all, sent by Postmark, SureContact or anything else, and every Supabase Auth email from one generator.
Bricks 2.4 and later, Automatic.css and SureCart: the WordPress stack, proven rather than assumed, and sign-in screens on brand, including the Fortress security plugin’s two-factor pages.
Cloudflare Pages, Stripe billing on Supabase, passkey sign-in on Supabase Auth, SureContact and OttoKit: the web-app stack, proven rather than assumed.
Campaign tracking set up the same way for every client: short links from a Google Sheet, first-touch UTM tags without cookies, leads and purchases sent to Meta from the server, and cost per lead in one sheet.
Every pack, now and later: web pages, several agents, email, WordPress, web apps and campaign tracking.
The six separately: $234