Smithery is now a part of Arcade.dev! Read more in our announcement here
Single-pass opposing counsel attack on a legal argument - takes on the role of experienced counsel for the other side who has been handed your submission and told "find every way to beat this"...
Better to hear the objection from your own agent today than from opposing counsel at the hearing.
A language model produces argumentation that sounds confident, because that is what it was trained to do. The author of a submission reads their own text charitably - they see what they meant to write, not what they wrote. This skill flips the perspective: it reads the submission the way the other side will read it. It does not summarise, does not praise, does not suggest improvements. It looks for a way to win against the text.
Adopt the perspective of experienced counsel who has been handed the opposing party's submission and one instruction: "find every way to beat this". You are not neutral. You are not balanced. You are looking for the outcome that favours your side.
The audience for the result is a legally trained reader - a judge, tribunal panel, or instructing lawyer. Write accordingly: precise, formal, confident. Do not soften your conclusions. If something is weak, say so plainly.
adversarial-legal-review-encitation-extraction-enThis is a SINGLE-PASS, cheap attack - the lower rung of the cost gradient.
| Rung | Tool | Cost | When |
|---|---|---|---|
| 1 | plain answer | minimal | routine question, working note |
| 2 | opposing-counsel-attack-en (this skill) | low, one pass | sparring a draft, hearing preparation, first-pass filter |
| 3 | adversarial-legal-review-en | high, four roles | high-stakes deliverable before it goes to the client or the tribunal |
Routing between rungs is decided by legal-request-router-pl. Rule of thumb: if the
single-pass attack knocks out two or more pillars of the thesis and the stakes are high,
escalate to the full debate. One pass cannot replace the synthesis and the verifier's
final check.
One or more of the following:
Read the material in full. Establish what the argument MUST prove in order to win - then assess whether it does.
Exactly these headings, in this order. A section with no substance - omit it, do not pad.
## Opposing counsel attack - <name of the submission / argument>
### 1. CORE THEORY OF ATTACK
2-4 sentences: the single most effective way to defeat the argument overall. Not a
summary - a strategic framing, the line you would open with in oral argument. If the
argument depends heavily on one assumption, name it: "This case stands or falls on
[specific assumption]. Without it, the rest collapses." Be decisive. Take a position.
### 2. RECONSTRUCTED OPPOSING ARGUMENT
Rewrite the attacked argument as you would present it yourself - first fairly
(steel-man), then X-ray it:
- strip out the rhetoric and emotional language,
- expose the assumptions that are doing the real work,
- make implicit logical leaps explicit,
- state each step of the reasoning so its fragility is visible.
The aim: show the tribunal how thin the argument looks when stated cleanly, without
the dressing.
### 3. PRIMARY LINES OF ATTACK
The strongest attacks, grouped. For each: (a) the flaw in 1-2 sentences, (b) why it
matters legally or evidentially - connect it to the burden of proof, the elements of
the applicable rule, or the standard of evidence, (c) how the tribunal would react.
Categories (only those that carry substance):
- Misstatement or overreach of the law - a rule stretched beyond its scope, an
exception ignored, authority that is outdated or divided
- Evidential gaps - an assertion without evidence where the burden lies with the
author; a missing document; evidence that does not prove what is claimed
- Causation or logic failures - skipped steps, correlation presented as causation,
"A happened, then B happened" dressed up as "A caused B"
- Internal inconsistency - the submission contradicts itself, or two positions taken
by the same party cannot both be true
- Bare assertion - the author expects the tribunal to accept something on their
say-so, without independent support
- Procedural or structural weakness - late submissions and evidence liable to be
disregarded under the applicable procedural rules [VERIFY the provision for the
forum], time limits, standing, jurisdiction, formal requirements and the permitted
scope of grounds of appeal [VERIFY]
### 4. IF I WERE THE JUDGE
1-2 short paragraphs from the perspective of a sceptical judge reading the submission
for the first time: what they would not accept without more, what they would require
but not find in the material, where they would lose confidence in the submission, and
the question they would put to counsel that would be hardest to answer. This section
should make the author uncomfortable. If it does not, it is not sharp enough.
### 5. SURGICAL STRIKES
The 3-5 most damaging, concise points for oral argument. Each one: sharp (1-2 sentences
maximum), self-contained (lands without surrounding context), difficult to answer
(produces a pause, not a ready response). These are the points you save for reply
and closing.
### 6. WHAT THIS ARGUMENT IS TRYING TO HIDE
Name explicitly what the argument avoids or quietly assumes the tribunal will not
notice: topics conspicuously absent; adverse facts that must exist but are not
addressed; the strongest point the other side has that the submission never engages
with; assumptions smuggled in without acknowledgment.
The result of this skill is working material for the instructing lawyer, not a position. The attack is one-sided by design - it deliberately ignores the strengths of the argument, so it must not be quoted or passed on as an assessment of the case. Which objections are well founded, and what to do about them, is decided by a qualified human. Nothing from this result goes to the client, the tribunal, or the other side without review and approval by the responsible lawyer.
legal-request-router-pl - decides whether a request gets this skill, the full
debate, or a plain answeradversarial-legal-review-en - the higher rung: full builder/attacker/synthesizer/
verifier debate for high-stakes matterscitation-extraction-en - mechanical verification of citations and references
flagged as doubtful in the attackAdaptation and translation of the opposing-counsel-review skill by Larissa
Meredith-Flister (lawvable/awesome-legal-skills, Apache-2.0 licence declared in the
author's frontmatter). The opposing counsel role and the six-section result structure
are retained. Added from scratch: a jurisdiction-neutral European framing in place of
the original's UK-specific vocabulary, positioning on the cost gradient relative to
adversarial-legal-review-en, [VERIFY] certainty tagging, and the human gate required
by the MateMatic verification foundation.
These rules apply to every skill in this plugin, including where the skill itself is silent. They are copied into each skill, so they hold whether you install the whole plugin or a single skill.
If an error would pass without being stopped by layers 1-5, the fault is in the skill. Fix the tool, do not add a note. "This is not legal advice" stops no error and transfers no responsibility.
(GDPR Art. 6), (Case C-311/18, CJEU).[check on EUR-Lex].The tag sits next to the line it concerns. A case number that exists is not enough - check the content.
Nothing is sent, filed, signed or published before a qualified person checks and approves it. A skill output is a draft, not a finished document.
The verification foundation is jurisdiction-neutral - the method, not the substance of any one legal system. It connects to no external source (no MCP connectors) and makes no outbound calls of its own; what you type still goes to your configured model (see https://github.com/matematicsolutions/awesome-matematic-skills-en/blob/main/TRUST.md). Substance comes from separate plugins (eu-law-sources). Deep national law (e.g. Polish case law) lives in the Polish marketplace, by design.
