Ready to supercharge your AI agents?

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

Get Started Free →

Rank Comparison · yogsoth-ai | DecimalAI

yogsoth-ai/rank-comparison

SkillScore78Maintained

Compare multiple ranking results to assess agreement and identify divergent items.

researchby yogsoth-ai

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

Automatic

Data Analyst pack

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

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-rank-comparison/SKILL.md

decimalai skills pull yogsoth-ai-rank-comparison --out .claude/skills/

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

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-rank-comparison/SKILL.md https://app.decimal.ai/s/yogsoth-ai-rank-comparison/SKILL.md

Raw file: /s/yogsoth-ai-rank-comparison/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-rank-comparison.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

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 18d ago

Yes — +22% lift · at higher cost+104% tokens0% turns78% of whole cases pass with skill (18/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-03	✗→✓	▲ Improved	270%	0%
case-04	✗→✓	▲ Improved	100%	0%
case-06	✗→✓	▲ Improved	409%	0%
case-13	✗→✓	▲ Improved	87%	0%
case-16	✗→✓	▲ Improved	160%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

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

yogsoth-ai / de-anthropocentric-research-engine

skills/rank-comparison/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/10/2026Version v1

Skill contents

SKILL.md

Rank Comparison

Compare ranking results produced by multiple methods or weight sets, compute consistency metrics, and identify alternatives with significant ranking differences.

Execution

Subagent receives multiple ranking results, computes rank correlation metrics, and outputs a consistency matrix and difference analysis.

Why Subagent

Rank comparison involves statistical computation (Kendall tau, Spearman rho) and difference attribution analysis; independent execution ensures computational accuracy.

HARD-GATE

Must report at least one rank correlation metric (Kendall tau or Spearman rho), and must list all alternatives with ranking differences >= 2 positions.

Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use | | --- | --- | | spawn-agent | Spawn a customized CC subagent with full MCP tool access. Used by SOPs that declare execution: subagent. |

Illustrative Want to see what yogsoth-ai/rank-comparison 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

9/3/2026, 22:30 UTC · 23 cases

Lift

+22%

78% vs 57%

Agent turns

0%

1.0 vs 1.0

Tokens

+104%

5,672.65 vs 2,777.7

fail → pass · 6pass → pass · 12fail → fail · 4pass → fail · 1

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-14 Analyst 1 and Analyst 2 published power rankings for 10 football teams. Analyst 1: [T1, T2, T3, T4, T5, T6, T7, T8, T9, T10]. Analyst 2: [T1, T3, T5, T2, T4, T7, T6, T9, T8, T10]. Provide a rank consistency report.	pass→pass	20,228	31,441	+55%	1	1	0%	3,212	6,187	+93%	0	0	—
case-01 Our team generated candidate shortlists using two distinct weighting schemes for our hiring criteria (Tech-heavy vs Leadership-heavy). Could you analyze the agreement between these two ranked lists? Please generate statistical correlation metrics between the runs and provide a summary table of candidates that show significant position shifts when changing the weights.	fail→fail	12,140	34,150	+181%	1	1	0%	2,070	6,126	+196%	0	0	—
case-02 I'm reviewing four different search relevance algorithms that each ranked the top 15 product results differently. I need an analytical report comparing these ranking sets. Specifically, output a correlation assessment showing pairwise agreement between the algorithms, and identify all product items that experience major rank movements across the different search models.	fail→fail	26,774	35,792	+34%	1	1	0%	4,173	8,451	+103%	0	0	—
case-19 Moderation Rule Set Alpha ranked 7 flagged posts as [Post1, Post2, Post3, Post4, Post5, Post6, Post7]. Rule Set Beta ranked them as [Post3, Post1, Post2, Post5, Post4, Post7, Post6]. Perform a statistical rank comparison.	pass→pass	22,887	24,897	+9%	1	1	0%	3,838	4,727	+23%	0	0	—
case-03 We evaluated 10 cloud vendors using two scoring frameworks (Framework A and Framework B). Framework A ranked Vendor Alpha at #1, Vendor Beta at #2, Vendor Gamma at #3, Vendor Delta at #4, Vendor Epsilon at #5. Framework B ranked Vendor Beta at #1, Vendor Delta at #2, Vendor Alpha at #3, Vendor Epsilon at #4, Vendor Gamma at #5. Most stakeholders just focus on top 1 movements, but we need an audit. Compare these two rankings and list any vendors that shifted positions.	fail→pass	7,378	28,587	+287%	1	1	0%	1,431	5,288	+270%	0	0	—
case-04 Our quantitative model produced a stock priority ranking based on momentum, while our fundamental model produced a ranking based on valuation. We want to see how well these models agree across our 8 target tickers. A colleague suggested reporting Pearson correlation and flagging only stocks moving at least 3 places. Perform the ranking comparison according to standard rank metrics.	fail→pass	24,692	41,631	+69%	1	1	0%	3,607	7,219	+100%	0	0	—
case-05 Random Forest feature importance ranked our 12 dataset features as [F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12]. XGBoost feature importance ranked them as [F2, F1, F5, F3, F4, F6, F8, F7, F9, F10, F12, F11]. Analyze the consistency between these two feature importance rankings and highlight discordant features.	pass→pass	21,104	29,549	+40%	1	1	0%	3,122	5,844	+87%	0	0	—
case-06 We tracked 15 target keywords on Google vs Bing. Our manager wants a quick summary and suggested listing only keywords that dropped or jumped 4+ spots. Provide a full rank comparison between the two search engine rank lists.	fail→pass	13,952	36,154	+159%	1	1	0%	1,606	8,177	+409%	0	0	—
case-07 An admissions committee compared holistic review scores against standardized test rank orders for 10 applicants. The preliminary analysis only looked at top-3 agreement. Provide a pairwise consistency matrix and difference audit for these applicant rankings.	pass→pass	57,929	39,664	-32%	1	1	0%	4,586	8,433	+84%	0	0	—
case-08 Model X (collaborative filtering) and Model Y (content-based) produced two top-10 movie recommendations for User 402. Model X order: [M1, M2, M3, M4, M5, M6, M7, M8, M9, M10]. Model Y order: [M3, M1, M2, M5, M4, M8, M6, M7, M10, M9]. Analyze the agreement between these two recommendation sequences.	pass→pass	18,273	34,092	+87%	1	1	0%	3,690	6,696	+81%	0	0	—
case-09 The QA lead ranked 8 critical bugs by user impact [B1, B2, B3, B4, B5, B6, B7, B8], while the Security lead ranked them by exploitability [B3, B1, B6, B2, B4, B5, B8, B7]. Evaluate the correlation between QA and Security priorities and identify discordant bugs.	pass→pass	19,192	22,085	+15%	1	1	0%	2,810	4,966	+77%	0	0	—
case-10 Risk Team Alpha ranked 6 key suppliers as [S1, S2, S3, S4, S5, S6]. Risk Team Beta ranked them as [S2, S4, S1, S3, S6, S5]. Provide a comparative report evaluating agreement and listing shifted suppliers.	pass→pass	19,032	27,676	+45%	1	1	0%	2,682	5,158	+92%	0	0	—
case-11 Three LLM benchmark runs evaluated 10 model checkpoints across coding tasks. Model Run 1, Run 2, and Run 3 produced slightly different leaderboards. Analyze the rank consistency across all three runs.	fail→fail	17,271	43,039	+149%	1	1	0%	2,068	8,434	+308%	0	0	—
case-12 Agent A ranked 7 neighborhoods as [N1, N2, N3, N4, N5, N6, N7]. Agent B ranked them as [N2, N1, N5, N3, N4, N7, N6]. Compare these two agent rankings.	pass→pass	16,894	18,718	+11%	1	1	0%	2,291	4,517	+97%	0	0	—
case-13 Enterprise clients ranked 8 upcoming features as [FeatA, FeatB, FeatC, FeatD, FeatE, FeatF, FeatG, FeatH]. SMB clients ranked them as [FeatC, FeatA, FeatB, FeatE, FeatD, FeatH, FeatF, FeatG]. Analyze the alignment between Enterprise and SMB feature rankings.	fail→pass	13,424	26,498	+97%	1	1	0%	2,628	4,913	+87%	0	0	—
case-15 Panel A ranked 6 climate proposals [P1, P2, P3, P4, P5, P6]. Panel B ranked them [P3, P1, P2, P6, P4, P5]. Analyze the agreement between Panel A and Panel B.	pass→pass	19,907	27,800	+40%	1	1	0%	3,017	4,081	+35%	0	0	—
case-16 Clinical Site Selection Team 1 ranked 8 potential hospital sites [H1, H2, H3, H4, H5, H6, H7, H8] based on patient volume. Team 2 ranked them [H2, H4, H1, H3, H6, H5, H8, H7] based on administrative speed. Compare these site rankings.	fail→pass	17,652	46,857	+165%	1	1	0%	2,464	6,400	+160%	0	0	—
case-17 Manager A ranked 6 team members [E1, E2, E3, E4, E5, E6]. Manager B ranked them [E1, E3, E5, E2, E4, E6]. Compare the agreement between Manager A and Manager B's rankings.	pass→pass	17,555	26,697	+52%	1	1	0%	2,683	4,974	+85%	0	0	—
case-18 Algorithm 'Best Match' vs Algorithm 'Top Rated' yielded rankings for 8 laptops: Best Match [L1, L2, L3, L4, L5, L6, L7, L8] vs Top Rated [L2, L1, L4, L3, L6, L5, L8, L7]. Someone recommended showing only items shifting by 3 or more places. Generate the full rank comparison report.	fail→pass	14,837	25,748	+74%	1	1	0%	2,045	4,851	+137%	0	0	—
case-20 Campaign Channel 1 ranked 8 ad creatives by click-through rate [C1, C2, C3, C4, C5, C6, C7, C8]. Campaign Channel 2 ranked the same creatives by conversion rate [C3, C1, C4, C2, C6, C5, C8, C7]. Provide a rank correlation analysis and identify discrepancies.	pass→pass	25,736	22,886	-11%	1	1	0%	4,109	5,300	+29%	0	0	—
case-21 We collected 50 data points recording annual marketing expenditure in thousands of dollars and annual revenue growth percentage. Calculate the correlation between marketing expenditure and revenue growth to evaluate linear association.	fail→fail	6,996	7,015	+0%	1	1	0%	1,234	422	-66%	0	0	—
case-22 We have raw performance scores for 5 server configurations across latency (ms), throughput (req/s), and cost ($/mo). Latency scores: [20, 15, 30, 25, 10]. Throughput scores: [1000, 1200, 800, 950, 1500]. Cost scores: [200, 250, 150, 180, 300]. Combine these scores using a weighted sum model with weights 0.4, 0.4, 0.2 to produce a single aggregated ranking.	pass→fail	21,304	40,800	+92%	1	1	0%	3,999	8,533	+113%	0	0	—
case-23 Write an SQL query to select all records from the customer table and sort them by total spend in descending order, with secondary sorting by signup date in ascending order.	pass→pass	7,801	8,815	+13%	1	1	0%	522	774	+48%	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. 23 cases were attempted. The headline lift of +22 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 cases got worse with the skill loaded, and they are 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: 56.5%
Pass rate with the skill: 78.3%
Cases attempted: 23
Cases counted toward lift: 23
Cases that regressed: 2
Skill version measured: v1
Run completed: 9/3/2026, 22:30 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-03

Prompt

We evaluated 10 cloud vendors using two scoring frameworks (Framework A and Framework B). Framework A ranked Vendor Alpha at #1, Vendor Beta at #2, Vendor Gamma at #3, Vendor Delta at #4, Vendor Epsilon at #5. Framework B ranked Vendor Beta at #1, Vendor Delta at #2, Vendor Alpha at #3, Vendor Epsilon at #4, Vendor Gamma at #5. Most stakeholders just focus on top 1 movements, but we need an audit. Compare these two rankings and list any vendors that shifted positions.

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 run78

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+69% 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 17 days ago

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