Ready to supercharge your AI agents?

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

Get Started Free →

Benchmark Inventory · yogsoth-ai | DecimalAI

yogsoth-ai/creative-ideation-benchmark-inventory

NewSkillScoreMaintained

Catalog all known solutions/methods in a domain with performance, applicability, and limitations.

researchby yogsoth-ai

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

Automatic

Research pack

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

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-creative-ideation-benchmark-inventory/SKILL.md

decimalai skills pull yogsoth-ai-creative-ideation-benchmark-inventory --out .claude/skills/

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

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-creative-ideation-benchmark-inventory/SKILL.md https://app.decimal.ai/s/yogsoth-ai-creative-ideation-benchmark-inventory/SKILL.md

Raw file: /s/yogsoth-ai-creative-ideation-benchmark-inventory/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-creative-ideation-benchmark-inventory.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 7d ago

—+26% tokens0% turns17% of whole cases pass with skill (4/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-20	✓→✓	＝ Same ✓	23%	0%
case-21	✓→✓	＝ Same ✓	-7%	0%
case-22	✓→✓	＝ Same ✓	16%	0%
case-23	✓→✓	＝ Same ✓	62%	0%
case-18	✗→✗	＝ Same ✗	41%	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 — 8/3/2026

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/creative-ideation-benchmark-inventory/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/3/2026Version v1

Skill contents

SKILL.md

Benchmark Inventory

Catalog all known solutions/methods in a domain. Produces a structured inventory with performance metrics, applicability scope, and known limitations.

Execution

Subagent — spawned via subagent-spawning/spawn-agent skill.

Why Subagent

Comprehensive benchmarking requires deep, focused research across multiple sources (papers, benchmarks, surveys). Benefits from dedicated context that can accumulate findings without polluting the orchestrator's state.

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. |

Want to see if yogsoth-ai/creative-ideation-benchmark-inventory would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/22/2026, 13:40 UTC · 23 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

0%

17% vs 17%

Agent turns

0%

1.0 vs 1.0

Tokens

+26%

5,593.43 vs 4,442.81

pass → pass · 4fail → fail · 19

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-18 Provide a comparative inventory of WebAssembly runtimes for edge computing. Detail startup latency metrics, memory footprint applicability, and platform system call limitations.	fail→fail	26,140	32,914	+26%	1	1	0%	3,901	5,483	+41%	0	0	—
case-05 I need a comparative benchmark catalog of consensus protocols used in permissioned distributed ledgers, such as Raft, PBFT, and HotStuff. Please list the established solutions along with throughput metrics under fault conditions, operational bounds, and vulnerability profiles.	fail→fail	27,928	36,965	+32%	1	1	0%	4,535	6,365	+40%	0	0	—
case-12 Build a structured inventory of end-to-end automated machine learning (AutoML) frameworks for tabular data. Include benchmark ROC-AUC scores, dataset row/column scale applicability, and search time limitations.	fail→fail	29,920	34,650	+16%	1	1	0%	5,117	5,707	+12%	0	0	—
case-01 I'm researching graph neural network architectures for link prediction. Could you build a comprehensive inventory of existing methods in this space, detailing each approach's benchmark metrics, practical application scope, and known failure modes or limitations?	fail→fail	30,907	12,289	-60%	1	1	0%	4,923	525	-89%	0	0	—
case-02 Can you catalog all current state-of-the-art approaches for zero-shot image segmentation? Please organize the findings into a clear breakdown showing evaluated performance figures, ideal context or use cases, and documented constraints for every technique.	fail→fail	27,705	28,745	+4%	1	1	0%	5,289	6,362	+20%	0	0	—
case-03 We need a thorough benchmark overview of post-training quantization techniques for large language models. Produce a structured list mapping out the known methods alongside their compression efficiency, target deployment scenarios, and inherent trade-offs or weaknesses.	fail→fail	33,833	32,523	-4%	1	1	0%	5,791	6,359	+10%	0	0	—
case-04 Our team is evaluating vector indexing algorithms like HNSW, IVF-PQ, and ScaNN for a high-throughput semantic search service. Provide a systematic inventory of candidate ANN indexing methods, detailing search latency and recall benchmarks, optimal dataset size bounds, and memory overhead limitations.	fail→fail	27,094	29,349	+8%	1	1	0%	4,577	5,425	+19%	0	0	—
case-06 We are selecting time-series anomaly detection algorithms for real-time IoT stream processing. Assemble an inventory of top algorithms in this domain, outlining benchmark accuracy and F1 scores, streaming rate applicability, and sensitivity to noise or missing data.	fail→fail	27,170	34,863	+28%	1	1	0%	4,483	5,274	+18%	0	0	—
case-07 Can you provide a structured benchmark overview of automatic speech recognition (ASR) fine-tuning approaches for low-resource languages? Detail word error rate metrics on benchmark corpora, hardware deployment applicability, and training data threshold limits.	fail→fail	23,221	34,031	+47%	1	1	0%	4,163	6,360	+53%	0	0	—
case-08 I am surveying continuous integration build caching strategies for large monorepos. Catalog the mainstream approaches, specifying build speedup metrics, monorepo scale limits, and failure recovery constraints.	fail→fail	26,412	37,824	+43%	1	1	0%	4,440	6,352	+43%	0	0	—
case-09 We need a comprehensive inventory of cross-platform mobile UI rendering frameworks. Break down performance frame rates, platform applicability, and native API access limitations.	fail→fail	31,806	33,218	+4%	1	1	0%	5,572	5,879	+6%	0	0	—
case-10 Please compile a benchmark catalog of distributed graph processing engines like GraphX, Giraph, and Galois. Detail throughput metrics on massive scale graphs, graph size applicability bounds, and memory spill constraints.	fail→fail	25,474	35,921	+41%	1	1	0%	4,224	6,354	+50%	0	0	—
case-11 Catalog modern serverless database solutions like Neon, CockroachDB Serverless, and PlanetScale. Provide cold-start latency metrics, transaction volume applicability scope, and connection pooling limitations.	fail→fail	30,937	38,492	+24%	1	1	0%	4,731	6,244	+32%	0	0	—
case-13 We are surveying privacy-preserving machine learning frameworks like PySyft and Concrete-ML. Produce an inventory listing encryption/execution overhead metrics, context applicability, and parameter size limitations.	fail→fail	23,861	15,388	-36%	1	1	0%	3,911	694	-82%	0	0	—
case-14 Catalog modern CSS-in-JS and zero-runtime styling solutions for React apps. Highlight bundle size metrics, server-driven rendering applicability, and dynamic styling limitations.	fail→fail	24,332	30,850	+27%	1	1	0%	4,063	5,264	+30%	0	0	—
case-15 Provide a benchmark inventory of modern distributed tracing collection agents. Detail CPU overhead metrics, target infrastructure applicability, and sampling loss limitations.	fail→fail	26,790	99,458	+271%	1	1	0%	4,656	6,341	+36%	0	0	—
case-16 We want a structured catalog of LLM agent orchestration frameworks like LangChain, AutoGen, and LlamaIndex. Map out agent execution latency, task complexity applicability, and state management constraints.	fail→fail	30,455	41,982	+38%	1	1	0%	4,270	6,178	+45%	0	0	—
case-17 Catalog modern database schema migration tools for cloud-native applications. Outline migration execution duration metrics, zero-downtime deployment applicability, and rollback constraints.	fail→fail	28,886	38,541	+33%	1	1	0%	4,698	6,346	+35%	0	0	—
case-19 Catalog Kubernetes service mesh implementations including Istio, Linkerd, and Cilium. Detail network latency overhead metrics, cluster size applicability, and eBPF kernel limitations.	fail→fail	38,127	38,063	-0%	1	1	0%	6,063	6,117	+1%	0	0	—
case-20 We are designing a new real-time collaborative whiteboarding feature for our web application. Brainstorm five creative architectural concepts for sync state handling and pitch the pros/cons of each ideation concept.	pass→pass	24,259	29,306	+21%	1	1	0%	3,545	4,362	+23%	0	0	—
case-21 Write a step-by-step Python code tutorial demonstrating how to quantize a Llama-3 model using AutoGPTQ INT4 post-training quantization.	pass→pass	16,354	13,966	-15%	1	1	0%	3,159	2,929	-7%	0	0	—
case-22 Explain the theoretical mathematical formulation behind the GraphSAGE aggregation mechanism and derive how node embeddings are updated during message passing.	pass→pass	22,211	23,946	+8%	1	1	0%	4,321	4,998	+16%	0	0	—
case-23 We are encountering a memory leak in our PyTorch training loop when calling loss.backward() in a custom loop. Help us debug and fix this memory leak in our code snippet.	pass→pass	9,839	14,721	+50%	1	1	0%	1,701	2,763	+62%	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, and 21 counted toward the lift figure. The other 2 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 0 percentage points is the difference between those two pass rates over the 21 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: 17.4%
Pass rate with the skill: 17.4%
Cases attempted: 23
Cases counted toward lift: 21
Cases that regressed: 2
Skill version measured: v1
Run completed: 8/22/2026, 13:40 UTC

SkillScore blends 4 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥20 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+18% tokens

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

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested 6 days ago

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