Ready to supercharge your AI agents?

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

Get Started Free →

Argument Crystallization · yogsoth-ai | DecimalAI

yogsoth-ai/argument-crystallization

SkillScore82Maintained

Distill the strongest arguments from each perspective through Argument Delphi or Dialectical Delphi methods.

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-argument-crystallization/SKILL.md

decimalai skills pull yogsoth-ai-argument-crystallization --out .claude/skills/

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

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

Raw file: /s/yogsoth-ai-argument-crystallization/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-argument-crystallization.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 8d ago

Yes — +64% lift · at higher cost+51% tokens0% turns82% of whole cases pass with skill (18/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	27%	0%
case-02	✗→✓	▲ Improved	30%	0%
case-03	✗→✓	▲ Improved	15%	0%
case-04	✗→✓	▲ Improved	77%	0%
case-05	✗→✓	▲ Improved	54%	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 — 8/3/2026

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/argument-crystallization/SKILL.md

View on GitHub ↗

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

Skill contents

SKILL.md

Argument Crystallization

Purpose: Rather than converging on a single answer, crystallize the strongest possible arguments for each position. Uses Argument Delphi (focus on argument quality over agreement) and Dialectical Delphi (thesis-antithesis-synthesis) to produce the most rigorous version of each stance.

When to use:

Policy deliberation requiring clear pro/con articulation
Interdisciplinary disputes where each field has valid concerns
Pre-decision analysis where decision-makers need best arguments
Situations where the goal is argument quality, not agreement

Budget

| Parameter | Constraint | |-----------|-----------| | Rounds | 2–3 (refine arguments, not opinions) | | Perspectives | ≥4 independent | | Argument quality gate | Each argument must be steel-manned |

State Ledger

| Key | Type | Description | |-----|------|-------------| | question | string | The deliberation question | | perspectives | array | Contributing perspectives | | initial_arguments | array | First-round arguments | | critiques | array | Cross-perspective critiques | | refined_arguments | array | Steel-manned final arguments | | synthesis | object | Points of agreement and irreducible tensions |

Available Tactics

disagreement-mapping — Identify argument clusters
iterative-convergence-round — Refine arguments across rounds

Available SOPs

judgment-collection
cluster-analysis
argument-extraction
feedback-distribution
consensus-measurement
consensus-synthesis

Execution Guidance

Collect initial positions with supporting arguments
Cross-distribute: each perspective critiques and steel-mans others
Authors refine arguments incorporating strongest critiques
Identify points of genuine agreement vs. irreducible tensions
Produce crystallized argument map with quality ratings

Output Format

yaml
positions:
  - label: <position name>
    strongest_arguments: [...]
    acknowledged_weaknesses: [...]
    steel_man_version: <best possible formulation>
agreements:
  - point: <shared conclusion>
    strength: <how robust>
irreducible_tensions:
  - between: [position_a, position_b]
    nature: <empirical/value/priority>
    why_irreducible: <explanation>

Available Tactics

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

| Tactic | When to use | | --- | --- | | disagreement-mapping | Map disagreement structure by collecting judgments, clustering opinions, extracting arguments per cluster, and visualizing fault lines. | | iterative-convergence-round | Execute one full Delphi round — collect judgments, distribute anonymous feedback, measure consensus, decide whether to continue. |

Available SOPs

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

| SOP | When to use | | --- | --- | | consensus-synthesis | Synthesize all rounds into a final consensus report documenting agreements, dissent, and process. |

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

Latest benchmark · publisher runVerified

8/22/2026, 03:19 UTC · 22 cases

Lift

+64%

82% vs 18%

Agent turns

0%

1.0 vs 1.0

Tokens

+51%

3,753.05 vs 2,478.64

fail → pass · 16pass → pass · 2fail → fail · 2pass → fail · 2

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-01 Our executive team is debating whether to mandate a three-day return-to-office policy or maintain full remote flexibility. We don't want a forced compromise—we need to see the absolute best versions of each argument. Please analyze this across multiple distinct organizational viewpoints. Output the result as a YAML structure detailing each position with its primary arguments, acknowledged weaknesses, and steel-manned formulation, followed by shared points of agreement and any irreducible tensions between conflicting views.	fail→pass	29,243	31,501	+8%	1	1	0%	4,133	5,267	+27%	0	0	—
case-02 We are conducting a policy review on whether open-weights AI models should be restricted by law due to biosecurity risks. Rather than picking a side, help us map out the strongest arguments across safety advocates, open-source developers, industry incumbents, and civil liberties groups. Present your findings in a YAML document containing each position's key arguments, admitted weaknesses, and best steel-man version, alongside any underlying consensus and irreducible value or empirical tensions.	fail→pass	24,225	27,540	+14%	1	1	0%	3,403	4,426	+30%	0	0	—
case-03 Our municipal council is deciding whether to implement congestion pricing in the city center. We need an analysis showing the most rigorous arguments from business owners, commuters, transit advocates, and environmental experts. Format the output in YAML listing the positions with their strongest supporting points, recognized flaws, and steel-man arguments, plus explicit sections for shared conclusions and irreducible trade-offs between positions.	fail→pass	23,288	22,598	-3%	1	1	0%	3,358	3,874	+15%	0	0	—
case-04 A state energy commission is weighing whether to extend the operational license of a nuclear power plant. Environmental advocates, grid operators, local residents, and climate economists disagree sharply. Format the evaluation in YAML showing each viewpoint's best case, admitted vulnerabilities, steel-manned thesis, shared points, and structural friction.	fail→pass	17,793	23,767	+34%	1	1	0%	2,365	4,192	+77%	0	0	—
case-05 Financial regulators are considering strict reserve requirements for algorithmic stablecoins. We need a YAML mapping of positions from retail investors, central bankers, crypto developers, and anti-money laundering enforcement officers. Avoid taking a side or picking a winner.	fail→pass	14,534	18,370	+26%	1	1	0%	2,126	3,273	+54%	0	0	—
case-06 A state legislature is debating whether gig economy drivers must be classified as full employees. Provide a YAML document mapping arguments from labor unions, platform companies, drivers desiring flexibility, and consumer advocates.	fail→pass	14,489	22,452	+55%	1	1	0%	2,129	3,554	+67%	0	0	—
case-07 A city planning department wants a comprehensive breakdown of arguments regarding mandatory single-family zoning abolition. Produce a YAML output detailing the stances of housing advocates, long-time homeowners, urban planners, and local school districts.	fail→pass	11,227	21,393	+91%	1	1	0%	1,762	3,836	+118%	0	0	—
case-08 Multinational corporations are wrestling with sovereign cloud mandates requiring local data storage. Model the debate across enterprise IT, national security agencies, privacy watchdog groups, and cloud infrastructure providers in YAML format.	fail→pass	18,127	19,642	+8%	1	1	0%	2,654	3,588	+35%	0	0	—
case-09 Transportation policy analysts are assessing strict liability vs negligence standards for autonomous vehicle accidents. Generate a YAML structure presenting arguments from automakers, insurance underwriters, personal injury attorneys, and safety advocates.	fail→pass	16,016	20,182	+26%	1	1	0%	2,554	3,500	+37%	0	0	—
case-10 Legal experts are divided on whether using copyrighted books to train large language models constitutes fair use. Synthesize arguments from visual artists, AI research labs, copyright scholars, and publishing executives into a YAML breakdown.	fail→fail	16,396	29,057	+77%	1	1	0%	2,586	4,746	+84%	0	0	—
case-11 International trade representatives are debating border carbon adjustments on heavy industrial imports. Produce a YAML map reflecting positions from domestic steel manufacturers, developing nation exporters, climate activists, and trade lawyers.	fail→pass	12,367	25,379	+105%	1	1	0%	1,849	4,413	+139%	0	0	—
case-12 Agricultural authorities are reviewing regulatory approvals for CRISPR-edited crops. Draft a YAML document detailing arguments from commercial farming conglomerates, organic farming associations, bioethicists, and consumer safety groups.	fail→pass	16,449	18,840	+15%	1	1	0%	2,859	3,311	+16%	0	0	—
case-13 The International Seabed Authority is considering a ten-year moratorium on deep-sea mineral extraction. Present a YAML summary covering mining consortiums, marine biologists, island nation governments, and clean energy battery manufacturers.	fail→pass	12,489	18,489	+48%	1	1	0%	2,069	3,366	+63%	0	0	—
case-14 A school district board is deliberating a total ban on student smartphones during school hours. Map out arguments from teachers, parents concerned about emergency contact, educational psychologists, and student representatives in YAML format.	pass→pass	12,656	20,606	+63%	1	1	0%	1,939	3,643	+88%	0	0	—
case-15 A regional government is reviewing whether to index the state minimum wage directly to consumer price index inflation. Produce a YAML crystallization covering small business owners, low-income labor advocates, macroeconomists, and retail trade groups.	fail→pass	16,978	22,481	+32%	1	1	0%	2,534	3,972	+57%	0	0	—
case-16 Lawmakers are drafting legislation to require public disclosure of social media recommendation algorithms. Generate a YAML breakdown capturing perspectives from tech platforms, child safety advocates, civil rights organizations, and cybersecurity researchers.	fail→pass	13,123	21,171	+61%	1	1	0%	1,966	3,868	+97%	0	0	—
case-17 A river basin authority must reconcile competing claims over declining river flow between agricultural irrigation districts, growing municipalities, indigenous tribes, and salmon restoration non-profits. Structure the analysis in YAML.	fail→pass	22,642	22,691	+0%	1	1	0%	3,450	3,854	+12%	0	0	—
case-18 An international space agency forum is negotiating mandatory de-orbiting deadlines for satellite constellations. Produce a YAML report analyzing stances from commercial satellite operators, defense agencies, astronomy researchers, and space insurance firms.	pass→pass	26,055	25,169	-3%	1	1	0%	3,945	4,066	+3%	0	0	—
case-19 A municipal government evaluated a two-year universal basic income trial and is now considering permanent implementation. Produce a YAML summary detailing positions from fiscal conservatives, anti-poverty advocates, labor economists, and local chamber of commerce representatives.	fail→pass	12,819	14,070	+10%	1	1	0%	2,027	2,697	+33%	0	0	—
case-20 We conducted a stakeholder vote on five potential office locations with 50 committee members. Here are the raw tally numbers: Location A (22 votes), Location B (15 votes), Location C (8 votes), Location D (5 votes), Location E (0 votes). Synthesize these votes and issue a final recommendation on which office location the company should sign a lease for.	pass→fail	9,793	16,787	+71%	1	1	0%	1,716	3,290	+92%	0	0	—
case-21 Our executive committee needs a definitive recommendation on whether to acquire Startup X for $50M. Analyze the financial metrics and risk factors provided, make a final go/no-go determination, and present a single unified action plan for the CEO.	pass→fail	17,561	25,855	+47%	1	1	0%	2,548	4,478	+76%	0	0	—
case-22 Sort the following 10 customer feedback comments into positive, negative, and neutral buckets. Provide a summary count for each bucket and list the main complaints mentioned.	fail→fail	4,083	4,416	+8%	1	1	0%	558	1,353	+142%	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. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 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: 18.2%
Pass rate with the skill: 81.8%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 2
Skill version measured: v1
Run completed: 8/22/2026, 03:19 UTC

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

Prompt

Our executive team is debating whether to mandate a three-day return-to-office policy or maintain full remote flexibility. We don't want a forced compromise—we need to see the absolute best versions of each argument. Please analyze this across multiple distinct organizational viewpoints. Output the result as a YAML structure detailing each position with its primary arguments, acknowledged weaknesses, and steel-manned formulation, followed by shared points of agreement and any irreducible tensions between conflicting views.

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 ≥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+31% 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 8 days ago

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