Research/Benchmarks
Research · Legal AI Evals

The Same Legal Skill Narrowed the Gap With Claude, But Did Not Close It

Our first ScaffoldBench showed the same model performs differently across scaffolds. The market asked a sharper question: was Claude’s lead really about the scaffold, or just its domain Legal Skill? So we gave MikeOSS the same Privacy Skills and re-ran the tasks to isolate one variable.

LN Labs Research·Published July 2026·7 min read

In our first ScaffoldBench, we tested a simple thesis: the same model can perform differently depending on the scaffold around it. We took one LLM, Opus 4.8, and ran it across three legal AI environments (Claude Chat, Cowork with Legal Plugin, and MikeOSS) on a 40-task legal eval focused on data protection and operational resilience.

The result confirmed the thesis: the same model produced different results across different scaffolds. But after publishing, we got useful feedback from the market.

Was Cowork ahead because of the full scaffold architecture, or just because it had a domain-specific Legal Skill?

01 · Recap

Why we ran a second ScaffoldBench

Some people pointed out that Cowork with Legal Plugin may have outperformed MikeOSS not because of the full scaffold architecture, but because it carried a domain-specific Legal Skill. That was a fair point, and a testable one. So we decided to run a follow-up experiment.

ScaffoldBench · Part 1

The first benchmark established that the scaffold around a model materially affects performance on legal tasks. The same Opus 4.8 scored differently in plain chat, in Cowork, and in MikeOSS.

Read the first ScaffoldBench
02 · Question

Which part of the scaffold matters most?

Part 1 showed that scaffold matters. The next question was narrower: which part of it? To answer that, we decomposed the scaffolds into their components, then focused this benchmark on just one.

01Tool calling
02SkillsThis test
03Memory & context persistence
04Sub-agent orchestration
05System prompt design
06Planning & verification loops

For this follow-up, we isolated the component the market had flagged: Skills, specifically the Privacy Skills that ship inside Claude’s Legal Plugin.

03 · Method

The experiment

We held everything constant except one variable. Same two scaffolds, same model, same evaluation dataset, then we added the same Privacy Skills to MikeOSS. The goal was not to prove which scaffold is universally better; it was to measure how much MikeOSS improves when it carries the same Skills as Claude’s Legal Plugin.

Model
Opus 4.8
Dataset
20 tasks
Domain
Privacy
Variable
Privacy Skills
Three configurations compared
BASELINE
MikeOSS
Open-source scaffold, no Skills
+ SKILLS
MikeOSS + Privacy Skills
Same scaffold, Claude's Privacy Skills added
REFERENCE
Cowork + Legal Plugin
Full Claude Cowork scaffold
04 · Result

Adding Skills narrowed the gap, but did not close it

Privacy Skills improved MikeOSS by +3.0 points overall. But Claude Cowork still led by around 5 points. The Skill mattered. It was not the whole story.

Overall leaderboard
Overall score · 20 tasks
Final results
1
Cowork + Legal PluginTop
Full Claude Cowork scaffold
82.7
2
MikeOSS + Privacy Skills+3.0
Skills added to the OSS scaffold
77.7
3
MikeOSS
Baseline · no Skills
74.7

Skills improved MikeOSS by +3.0 points. Claude Cowork stayed ahead by about 5.

05 · Where it helped

The Skill helped most on reasoning

The most important signal was not the overall score. It was where the improvement happened. Privacy Skills had by far the strongest impact on how the model reasoned through privacy-specific legal tasks.

Baseline vs. + Skills
MikeOSS · by dimension
Evidence
Baseline
76
+ Skills
77
Change
+1.0
Legal soundness
Baseline
78
+ Skills
80
Change
+2.0
ReasoningBiggest gain
Baseline
70
+ Skills
76
Change
+6.0

This suggests the Privacy Skills did not mainly improve document finding or evidence extraction. They improved the way the model reasoned through privacy tasks. A good legal Skill is not just a folder of legal knowledge. It can shape how the model approaches the task, what risks it looks for, and how it structures the analysis.

06 · Relevance

Skills work best when they match the task

On tasks with high Skill-relevance, MikeOSS with Privacy Skills came much closer to Claude Cowork, a near-dead heat. This is probably the strongest result from the follow-up.

High Skill-relevance tasks
Overall score · matched tasks
Cowork + Legal Plugin
83.8
MikeOSS + Privacy Skills+6.8
83.1
MikeOSS
Baseline · no Skills
76.3

Domain-specific Skills can make a substantial impact when they are properly matched to the legal task. This supports the point OpenAI recently made about evaluations. Domain-specific scaffolding can help pull more capability out of the model.

07 · Cost

Same improvement, without losing the cost edge

Adding Privacy Skills raised MikeOSS cost by about 15%, but it stayed roughly 57% cheaper than Claude Cowork. In real legal workflows, cost per task is an evaluation criterion too.

Cost per task
USD · averaged over the eval set
Lower is cheaper
MikeOSS
$0.91
MikeOSS + Privacy Skills+15%
$1.05
Cowork + Legal Plugin
$2.42

If a system can improve performance while keeping a material cost advantage, that is an important design signal. Quality is not the only criterion that matters.

08 · Trade-off

Why the cost difference remained

To understand the gap, we looked deeper into the scaffold architecture. Cowork has more built-in capability; MikeOSS is more lightweight. Both facts show up across the same components.

Scaffold comparison
System prompt
Cowork + Legal Plugin

Broader prompt with more general agent instructions

MikeOSS + Privacy Skills

Smaller, focused prompt concentrated on the legal task

Verification loops
Cowork + Legal Plugin

More likely to run verification steps, todo lists, extra checks

MikeOSS + Privacy Skills

More direct execution, with fewer built-in verification loops

Sub-agent orchestration
Cowork + Legal Plugin

Splits tasks across sub-agents to reduce context dilution

MikeOSS + Privacy Skills

Lighter execution without the same orchestration overhead

Tool calling
Cowork + Legal Plugin

Wider variety of tool calls and more flexibility

MikeOSS + Privacy Skills

More constrained tool calling, including a cap on calls

Retrieval style
Cowork + Legal Plugin

Expansive; sometimes pulls larger amounts of document context

MikeOSS + Privacy Skills

Targeted document search and clause-level lookups

Output length
Cowork + Legal Plugin

Allows longer outputs, supporting more detailed reasoning

MikeOSS + Privacy Skills

Harder output ceiling, which forces more concise answers

Memory & context
Cowork + Legal Plugin

More advanced memory and context persistence

MikeOSS + Privacy Skills

More task-focused, less reliant on persistent context

More scaffold capability does not automatically mean better performance for every task. Cowork can run extra checks, use more tools, and produce longer reasoning, useful for complex or ambiguous work, but it also raises token spend. For these tasks, some of that spend did not materially improve the final result. Sometimes the best scaffold is not the most complex one. It is the one with the right level of capability for the task.

09 · Conclusion

What we learned

We expected that adding the same Privacy Skills to MikeOSS might almost fully close the gap with Claude Cowork. It did not, but it narrowed it in a meaningful way. The strongest improvement was in reasoning, and the best result appeared where the Skills were directly relevant to the tasks. At the same time, MikeOSS stayed materially more cost-efficient.

Domain-specific Skills are a major performance driver inside legal AI scaffolds, especially for the tasks they are designed for. But Skills alone do not explain the entire gap.

The broader scaffold architecture still matters, especially for evidence handling, verification, and consistency.

For legal AI teams, this reframes the question. Not only which model should we use and not even only which scaffold, but:

Which scaffold components actually improve performance for this specific legal workflow, and at what cost?

Where we go next

The future of legal AI will not be only about stronger models. It will also be about better legal Skills, better scaffold architecture, and better evaluation infrastructure to understand what actually works. That is the direction we want to keep exploring with ScaffoldBench.

Work with us

Legal AI needs benchmarks that measure Skills, scaffolds, and cost, not just models.

If you are building legal AI evals, legal-agent workflows, or applied AI systems for law firms and legal teams, we would be happy to compare notes.

Book a Discovery Call