Pre Motion Andrew Bird Agent Skill

Pre-Motion — adversarial premortem for UK litigation

You think you've built the strongest version of your case. Pre-Motion runs it through a structured adversarial pipeline to find where it actually loses — the procedural, substantive, evidentiary, and strategic failure modes opposing counsel will pull on first. The opposite of confirmation bias, by design.

For: solicitors stress-testing before issue, in-house counsel before sign-off, mediators valuing settlement, litigation funders pricing a matter, anyone deciding whether to take a case.

How it runs

Four passes over the matter. Run the four adversarial passes in Stage 3 as parallel sub-agents if your environment supports them; otherwise run them in sequence — the method and the output are the same either way.

• Optimistic baseline. Build the strongest version of the case the evidence supports. This is the foil for everything that follows.
• Evidence inspection. Three checks: document review (gaps, weak documents), cross-reference (one document contradicting another), chronology (timeline gaps, dates that don't fit). Produce evidence flags with a severity each.
• Premortem adversary. Four adversarial passes, one per failure category below. Give each the same frame: *"It is [trial date + 1 year]. This case has been LOST. Walk back — what in your category caused the loss?"* Produce ranked failure scenarios per category.
• Synthesis. Diff the optimistic baseline against the adversarial findings. Produce the brief.

--depth=fast runs a single combined adversarial pass for a quick read; --depth=thorough runs all four passes in full.

Inputs

• Matter facts: parties, brief chronology, claim heads, jurisdiction, forum.
• Evidence references: documents, witness statements, expert reports — pointers to matter content, not re-uploaded.
• The strongest version of the case as the user sees it (the optimistic baseline).
• Optional: the counterparty's pleaded or anticipated defence.
• Optional: --depth=fast|thorough.

Step 1 — Permitted-use check (CPR 31.22 + privilege)

Before reading matter documents:

• The host workspace enforces the hard gate (matter-slug match against the proceedings reference, privilege posture). If this skill is running, that gate has already passed — this step does not replace it.
• Still confirm with the user: are any documents drawn from disclosure in *other* proceedings? If so, stop until permission, the parties' agreement, or open-court reference is established (CPR 31.22 implied undertaking).
• If the matter's privilege posture is mixed, add a [PRIVILEGE FLAGGED] banner to the output and recommend counsel review before any external distribution.

Failure-mode categories

The four Stage 3 passes specialise in the four ways UK civil cases lose.

Procedural

• Limitation expired or contested (Limitation Act 1980, s.5 / s.11 / s.14A).
• Pre-action protocol non-compliance (Pre-Action Conduct PD, sector protocols).
• Strike-out or summary-judgment vulnerability (CPR 3.4, CPR 24).
• Costs sanctions risk (CPR 44.2).
• ADR refusal exposure (Halsey; Churchill v Merthyr Tydfil [2023]).
• Service / jurisdiction defects (CPR 6).
• Disclosure-regime missteps (CPR 31 vs PD 57AD).

Substantive

• Cause-of-action elements unproven.
• Causation gaps, factual or legal (Wagon Mound, Fairchild, SAAMCO).
• Mitigation failures (British Westinghouse).
• Affirmative defences (estoppel, waiver, release, contributory negligence, ex turpi causa).
• Statutory bars (Consumer Rights Act, UCTA, statutory limitation).

Evidentiary

• Privilege exposure (Unilever exceptions, joint-defence breakdown, Rush & Tompkins boundary).
• Disclosure failures and adverse-inference risk (CPR 31, PD 57AD).
• Witness credibility, availability, inconsistency.
• Hearsay weaknesses (Civil Evidence Act 1995 s.2–4 notice failures).
• Expert-report deficiencies (CPR 35; joint-instruction failures; Toth v Jarman).
• Document authenticity and chain of custody.

Strategic

• Settlement leverage misjudged (BATNA gap to the opposing side).
• Cost/benefit ratio misaligned with client objectives.
• Reputational or regulatory exposure from issue or trial.
• Information asymmetry working against the client.
• Counterparty's BATNA stronger than the optimistic baseline assumes.

Each pass should cite the relevant authority for its category, and mark any rule or case it cannot pin to a section or citation with [CITE NEEDED] rather than stating it as settled.

Output

Produce the brief with the sections below. Render it as the finished brief — do not echo this list back as a template, and do not invent facts to fill a section; if a section has nothing in it, say so.

• A reviewer-note line: *work product, prepared in contemplation of litigation, subject to litigation privilege.*
• A header: matter name, date generated, depth, privilege posture, and the verdict (Steelman / Borderline / Strawman).
• The one brutal sentence — "If we lose this, this will be why: [single sentence]."
• Optimistic baseline — the strongest version of the case, as the baseline pass built it.
• Ranked failure scenarios — grouped Procedural / Substantive / Evidentiary / Strategic. Each scenario is one paragraph with a Severity (H/M/L), a Likelihood (H/M/L), and a Mitigation.
• Evidence inconsistencies — the flags from Stage 2.
• Blind spots — issues the baseline assumed resolved that the adversary found open.
• Mitigations — one concrete action per scenario where applicable: strengthen evidence, amend pleadings, settle, withdraw, brief counsel differently.
• Settlement-posture implications — qualitative only. Translate the failure profile into posture, e.g. "several procedural failure modes ranked H — settle harder than the baseline suggests; consider a Calderbank offer before issue." Produce no number; full BATNA / ZOPA analysis belongs to the v0.2 settlement-helper skill.

Mark uncertainty inline as you go:

• [SME VERIFY — failure mode] — borderline adversary output; counsel's call.
• [CITE NEEDED — authority] — a rule or doctrine referenced without a section or case; verify before relying on it.
• [EVIDENCE FLAG — severity] — surfaced by Stage 2; check against the source document.

What this skill does not do

• Predict the outcome. It surfaces failure modes; outcomes depend on the tribunal, the judge, the witnesses, the day.
• Take the case for you. The verdict (steelman / strawman) is the model's read of the brief, not advice.
• Replace counsel's strategic call. Settle, withdraw, strengthen — all counsel decisions.
• Cover non-UK procedure (US federal, Scotland, NI).
• Run during trial. This is pre-action, pre-settlement, or pre-funding use.
• Replace a formal counsel opinion. A KC's view on case strength matters more than this output. Pre-Motion is a structured prompt for that conversation, not a substitute for it.

v0.2 roadmap

A separate settlement-helper skill covering Calderbank / Part 36 mechanics and BATNA / ZOPA / Nash bargaining — the quantitative settlement analysis deliberately kept out of this skill.