Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Verify This · kunanonj | DecimalAI

kunanonj/cursor-plugin-teamkit-verify-this

SkillScore82Maintained

Verify a claim with fresh local evidence: restate it falsifiably, capture baseline and treatment, compare artifacts, and return VERIFIED, NOT VERIFIED, or INCONCLUSIVE.

Testingby kunanonj

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

QA & Test Engineering pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/kunanonj-cursor-plugin-teamkit-verify-this/SKILL.md

decimalai skills pull kunanonj-cursor-plugin-teamkit-verify-this --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/kunanonj-cursor-plugin-teamkit-verify-this/SKILL.md https://app.decimal.ai/s/kunanonj-cursor-plugin-teamkit-verify-this/SKILL.md

Raw file: /s/kunanonj-cursor-plugin-teamkit-verify-this/SKILL.md ↗ · machine-readable summary at /s/kunanonj-cursor-plugin-teamkit-verify-this.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 49d ago

—-23% tokens0% turns82% of whole cases pass with skill (18/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	4%	0%
case-03	✗→✓	▲ Improved	55%	0%
case-04	✗→✓	▲ Improved	-32%	0%
case-05	✗→✓	▲ Improved	3%	0%
case-10	✗→✓	▲ Improved	-22%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/2/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

kunanonj / ai-skills-hub

skills/cursor-plugin-teamkit-verify-this/SKILL.md

View on GitHub ↗

License MITImported 8/2/2026Version v1

Skill contents

SKILL.md

Verify This

Verification is not a recap. It proves or disproves a specific claim with repeatable evidence.

When To Use

The user asks "verify this", "prove it works", "did this fix it", or "show me the evidence".
A bug fix needs a before/after repro.
A UI, CLI, API, performance, or memory claim needs measurement.
A test passes but the user-visible behavior still needs confirmation.

Do not use this for vague claims like "the code is cleaner". Ask for a measurable claim first.

Workflow

Restate the claim in falsifiable form: condition, metric, and threshold.
Pick the smallest local surface that can disprove it.
Capture a baseline from the old state: merge base, parent commit, failing branch, or current broken repro.
Capture treatment from the changed state with the same command, data, warmup, and environment.
Compare raw artifacts: numbers, screenshots, terminal transcripts, HTTP responses, profiles, heap snapshots, or test output.
Return exactly one verdict: VERIFIED, NOT VERIFIED, or INCONCLUSIVE.

Local Surfaces

Code behavior: focused unit/integration tests or a minimal repro script.
CLI/TUI behavior: control-cli, terminal transcript, or demo recording.
UI behavior: control-ui, screenshots, accessibility snapshots, or browser traces.
API behavior: local HTTP/RPC request and response diff.
Performance: same-machine baseline/treatment timings or CPU profiles.
Memory: heap snapshots before and after the suspected operation.

Artifact Layout

When safe to write artifacts:

text
/tmp/verify-this/<claim-slug>/
├── claim.md
├── timeline.md
├── baseline/
├── treatment/
├── diff/
└── verdict.md

If artifacts may contain sensitive code, prompts, screenshots, HTTP bodies, or heap data, keep only the minimal inline evidence unless the user agrees to disk storage.

Verdict Rules

VERIFIED: baseline and treatment differ in the predicted direction, by the claimed threshold, with no obvious confound.
NOT VERIFIED: the behavior is unchanged, moves the wrong way, or misses the threshold.
INCONCLUSIVE: no valid baseline, noisy signal, failed measurement, or an environment difference invalidates the comparison.

Output

Use this shape:

text
VERIFIED | NOT VERIFIED | INCONCLUSIVE
Claim: <falsifiable claim>

Evidence:
<metric/artifact>: baseline=<...>, treatment=<...>, delta=<...>, threshold=<...>

Reasoning:
<one tight paragraph naming the evidence and any confounds>

Do not soften a negative result. A clear NOT VERIFIED is useful.

Illustrative Want to see what kunanonj/cursor-plugin-teamkit-verify-this would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/9/2026, 20:57 UTC · 22 cases

Measured on 21 comparable cases — below the 22 needed to publish a lift, so this result is shown here but not on the registry card or leaderboard.

Lift

+59%

82% vs 23%

Agent turns

1.0 vs 1.0

Tokens

-23%

1,526.95 vs 1,993.43

fail → pass · 14pass → pass · 4fail → fail · 3pass → fail · 1

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-22 Our production server keeps crashing with `java.lang.OutOfMemoryError: Java heap space` in `OrderProcessor.java`. Explain what typical causes lead to this error and how to debug JVM heap dumps.	pass→pass	15,545	15,005	-3%	1	1	0%	2,798	3,198	+14%	0	0	—
case-01 I just optimized our database query function in `services/db.js` to reduce memory usage under load. Can you verify this claim? Please format the output with a single verdict line, the restated claim, a baseline and treatment evidence summary highlighting the delta, and a brief reasoning paragraph.	fail→pass	8,846	5,905	-33%	1	1	0%	1,586	1,649	+4%	0	0	—
case-02 We applied a patch to `api/user.go` so HTTP 500 errors on invalid email registration are replaced by 400 Bad Request responses. Did this fix it? Provide a structured evaluation showing the verdict status, the claim under test, baseline versus treatment measurement evidence, and a short paragraph detailing your reasoning.	fail→fail	8,367	4,709	-44%	1	1	0%	1,511	898	-41%	0	0	—
case-03 I updated the CLI parser in `src/cli.ts` so execution completes in under 50ms. Show me the evidence to prove whether it works. Please structure your final output with a verdict line at the top, the falsifiable claim, an evidence breakdown comparing before and after results against the target, and a concise reasoning block.	fail→pass	5,867	9,158	+56%	1	1	0%	1,184	1,839	+55%	0	0	—
case-04 Our frontend team claims the checkout modal in `CheckoutModal.tsx` now opens in under 100ms. We ran browser trace snapshots before and after the change: baseline frame render was 180ms, treatment frame render is 65ms. Evaluate this verification request and present the verdict, restated claim, evidence breakdown, and reasoning.	fail→pass	11,622	4,180	-64%	1	1	0%	2,196	1,486	-32%	0	0	—
case-05 The backend team claimed the new Redis cache in `services/cache.go` increased GET /items throughput to over 5000 req/sec. In load testing with identical concurrency, baseline without Redis was 1200 req/sec and treatment with Redis reached 4800 req/sec. Formally evaluate this claim and present the verdict, claim details, evidence diff, and reasoning.	fail→pass	7,521	4,297	-43%	1	1	0%	1,553	1,595	+3%	0	0	—
case-06 We replaced `lodash` with native methods in `src/utils.ts` to shrink the minified bundle size by at least 15KB. Build output artifacts show baseline bundle size was 240KB and treatment bundle size is 220KB. Output a formal verdict report with verdict, claim, evidence breakdown, and concise reasoning.	fail→fail	3,962	4,327	+9%	1	1	0%	778	1,478	+90%	0	0	—
case-07 A developer claims fixing event listener cleanup in `WorkerPool.js` eliminated memory growth during job execution. Heap snapshots show baseline memory grew by 45MB over 1000 jobs, while treatment memory grew by 42MB over 1000 jobs. Evaluate this claim and render the result using the structured report format.	pass→pass	11,495	4,123	-64%	1	1	0%	2,013	1,417	-30%	0	0	—
case-08 We updated `NavigationMenu.tsx` to fix screen reader accessibility so all interactive elements have accessible names. An automated DOM audit shows baseline had 4 unlabelled buttons and treatment has 0 unlabelled buttons. Evaluate this verification request and structure the output as required.	pass→pass	8,906	3,556	-60%	1	1	0%	1,480	1,278	-14%	0	0	—
case-09 The terminal team updated `logger.py` so warning logs are colorized yellow on stderr. Terminal recording transcripts show baseline stderr output had plain text logs while treatment stderr contains ANSI color escape sequences `\033[33m` for warnings. Evaluate this claim and return a formal evaluation summary.	pass→pass	10,049	5,117	-49%	1	1	0%	1,840	1,639	-11%	0	0	—
case-10 An engineer claims fixing race condition in `auth_test.go` lowered test runtime from 8 seconds to under 2 seconds. Running the test suite locally gave baseline=8.1s, treatment=1.9s on machine A, but baseline=2.1s and treatment=7.8s on machine B due to background CPU contention. Produce the evaluation report for this claim.	fail→pass	13,690	5,416	-60%	1	1	0%	2,287	1,794	-22%	0	0	—
case-11 We optimized `calculateTax()` in `billing.ts` to reduce CPU time spend during checkout to under 5% of total request execution. pprof profiles show baseline CPU allocation was 14% and treatment CPU allocation is 3.2%. Provide the full evaluation output.	fail→pass	8,772	3,479	-60%	1	1	0%	1,632	1,333	-18%	0	0	—
case-12 We want to verify that `user_export.py` correctly obfuscates user passwords in JSON exports. We have generated export files containing user passwords in cleartext in the baseline and hashed strings in treatment. Provide the evaluation report, taking into account safe handling of sensitive output artifacts.	fail→pass	10,726	5,706	-47%	1	1	0%	2,000	1,575	-21%	0	0	—
case-13 We modified `controllers/orders.js` so missing authentication headers return HTTP status 401 instead of HTTP 403. Curl responses show baseline HTTP status was 403 Forbidden and treatment HTTP status is 401 Unauthorized. Provide the evaluation report with verdict, claim, evidence breakdown, and reasoning.	fail→pass	7,571	4,081	-46%	1	1	0%	1,434	1,480	+3%	0	0	—
case-14 A PR in `db/pool.ts` claims to cap maximum active database connections at 20 under 100 concurrent requests. Load test metrics show baseline peak connections reached 100 and treatment peak connections reached 19. Structure the formal verification evaluation.	fail→pass	19,057	2,968	-84%	1	1	0%	3,187	1,182	-63%	0	0	—
case-15 We enabled gzip compression on gRPC endpoints in `server.go` to cut network payload size by at least 50%. Network captures show baseline response size was 120KB and treatment response size is 45KB. Generate the verification evaluation.	fail→pass	5,842	3,442	-41%	1	1	0%	1,022	1,300	+27%	0	0	—
case-16 The serverless team claims removing heavy SDK imports reduced AWS Lambda cold start duration to under 500ms. CloudWatch metrics show baseline cold start was 1200ms and treatment cold start is 680ms. Provide the verification output.	fail→pass	7,172	3,366	-53%	1	1	0%	1,333	1,238	-7%	0	0	—
case-17 A developer claims optimizing CSS in `styles/main.css` brought First Contentful Paint (FCP) below 1.0 second on mobile devices. Lighthouse runs report baseline FCP was 2.4s and treatment FCP is 0.8s under identical throttling settings. Formally evaluate this claim.	fail→pass	16,768	4,077	-76%	1	1	0%	2,733	1,424	-48%	0	0	—
case-18 We updated `cache_manager.py` to pre-warm user session data, claiming cache hit ratio would exceed 90% during peak traffic. Metrics show baseline hit ratio was 65% and treatment hit ratio is 88%. Evaluate this claim with the required format.	fail→pass	8,099	3,232	-60%	1	1	0%	1,404	1,211	-14%	0	0	—
case-19 We switched from `golang:1.21` to `golang:1.21-alpine` in `Dockerfile` to shrink image size below 100MB. `docker images` output shows baseline size was 850MB and treatment size is 92MB. Evaluate this claim.	fail→pass	16,347	4,596	-72%	1	1	0%	3,032	1,551	-49%	0	0	—
case-20 Can you review `services/payment.js` and refactor it so the code looks cleaner, more readable, and follows SOLID principles?	fail→fail	1,354	7,073	+422%	1	1	0%	199	1,581	+694%	0	0	—
case-21 Write a complete suite of Jest unit tests for the user authentication helper functions in `src/utils/auth.ts` to ensure full line coverage.	pass→fail	26,308	3,748	-86%	1	1	0%	6,171	818	-87%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +59 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 22.7%
Pass rate with the skill: 81.8%
Cases attempted: 22
Cases counted toward lift: 21
Cases that regressed: 1
Skill version measured: v1
Run completed: 8/9/2026, 20:57 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

I just optimized our database query function in `services/db.js` to reduce memory usage under load. Can you verify this claim? Please format the output with a single verdict line, the restated claim, a baseline and treatment evidence summary highlighting the delta, and a brief reasoning paragraph.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run82

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns-7% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 49 days ago

Other measured skills in the registry, with their headline benchmark lift.