Home · Try the Engine · FRE 707 Disclosure · Actor Review Memo
Layout Preview

Actor Review Memo — See the Format Before You Generate

CPP's claim workbench produces a one-page analyst follow-up review memo for each recurring actor (3+ distinct slips) surfaced by the cross-reference engine. This page renders an example so you see the layout live, without uploading anything.

Synthetic actor + synthetic findings shown below. No real claim data is exposed. Replace each [FILL IN] with case-specific facts. This format is for analyst follow-up review. It is NOT a deposition prep tool; final witness preparation is for legal counsel.

Forensic disclosure
This format is for analyst follow-up review. The memo summarizes analyst-observable patterns from email NER + slip cross-reference. It does NOT establish fault, intent, or evidence tampering. It does NOT substitute for legal analysis or expert testimony preparation. It IS suitable for analyst follow-up review only.
Layout Key

How to read the preview below

The example memo below shows the exact layout produced by actor_review_memo_docx.py and its PDF sibling. Each section maps one-for-one to the Word document the workbench generates. Amber tokens flag fields the analyst must replace before sharing the memo.

Legend
Amber token — analyst-fillable placeholder Navy #0F2540 — structural elements (headings, rules) Red #C8392F — accent and citation markers
Example Memo — Single Page Per Actor cpm-engine v2.8.0 · method_id: actor_review_memo_v1

=== ACTOR REVIEW MEMO ===

CPP v2.8.0
computed: 2026-05-10T14:33:08+00:00
Actor: M. Sample (synthetic example)
Email provenance: Owner — analyst-supplied domain mapping (correlation, not affiliation) [72.4% corpus share] [1]
Distinct slips: 5 · Evidence items: 14 · Role tag: Project director

CONTEXT

This actor appears in 5 distinct delay slips across 3 schedule window(s). Communication trail spans 2025-09-02 to 2026-02-18.

KEY OBSERVATIONS

  • Slip 1: [Window W2, Activity A1240 [CP]] — Foundation cure milestone (12d slip) [2]
  • Slip 2: [Window W2, Activity A2105] — Underground utilities (7d slip) [2]
  • Slip 3: [Window W3, Activity A1240 [CP]] — Foundation cure milestone (additional 6d slip) [2]
  • Slip 4: [Window W3, Activity A3015] — Steel delivery acceptance (4d slip) [2]
  • Slip 5: [Window W4, Activity A4020 [CP]] — Roofing dry-in (9d slip) [2]

PARTY CLASSIFICATION

Email-domain-derived: Owner (72.4% corpus dominance) [1]. Classification uses analyst-supplied domain_party_mapping; not a determination of party-side actually responsible.

CONTESTED-SLIP OVERLAP

This actor appears in 2 contested slip(s) where party attribution is split. See contested-slips DOCX for full analysis. [2]

ANALYST FOLLOW-UP PROMPTS (NEUTRAL — NOT INVESTIGATIVE CONCLUSIONS)

  • Review the 14 email(s) from this actor in the corpus for context. [1]
  • Cross-reference dates against scheduled milestone meetings. [2]
  • Confirm the actor's role and authority on the project.
  • Identify whether the actor's communications align with notice/response obligations under [FILL IN: notice framework]. [3]

CITATIONS

  • [1]Email corpus excerpts — 14 evidence item(s) where this actor appears (subject/date/excerpt available in workbench HTML dashboard).
  • [2]Slip register rows — 5 distinct slip(s) paired with this actor; full per-slip data in workbench_result.cross_reference.pairings.
  • [3]Jurisdiction case-law ([FILL IN: jurisdiction]): analyst to consult notice-failure precedent for this jurisdiction.
Forensic Disclosure
This memo summarizes patterns from analyst-supplied evidence. It does NOT:
  • Establish fault, intent, or evidence tampering
  • Substitute for legal analysis or expert testimony preparation
  • Constitute a representation about the actor's conduct
It IS:
  • A neutral analyst-review document highlighting recurring patterns
  • Suitable for analyst follow-up review
  • Mathematically verifiable (topology hash: a3f9e2d18bc4…)
Per CPP forensic-correctness discipline: no LLM was used to generate this memo. Every claim cites a source row.
Prepared by: [FILL IN: analyst signature]
CPP cpm-engine v2.8.0 · computed_at 2026-05-10T14:33:08+00:00
Like the format? The memo is generated from your workbench result via actor_review_memo_docx.build_actor_review_memo_docx(). Every recurring actor (3+ distinct slips) gets a one-page Word document. The same data drives the PDF sibling.
Back to /try →
Web generation deferred to v0.2. The DOCX/PDF generators live in claim-workbench/scripts/actor_review_memo_docx.py and are wired into the local claim workbench skill (run from a workbench result on disk). A web endpoint (POST /api/generate-actor-memo) that accepts a JSON workbench-result payload is on the v0.2 roadmap. For now, run the workbench locally and the memos are written to your output directory automatically.
Read this before you share the memo

Forensic-neutral by construction. Analyst follow-up only.

CPP's claim-workbench memos use neutral verbs by design. They surface patterns. They flag follow-up questions for the analyst. They do not draw conclusions about an actor's conduct, fault, or intent.

Neutral verbs

"Appears", "aligns with", "may warrant review".

Every conclusory verb is hedged. The memo never says "is responsible", "caused", or "failed to". The forbidden-phrase blocklist is enforced by a regression test on every generated output.

Correlation, not affiliation

Email-domain provenance is a correlation only.

Party classification uses analyst-supplied domain_party_mapping. The tool surfaces the dominant party label and corpus share. It never claims who an actor is "affiliated with" or "working for" — just whose domains they correspond from.

Citation-anchored

Every numerical claim anchors to a bracketed [N] citation tag.

The closing citations block is the source map. Email-corpus counts cite [1]. Slip-register counts cite [2]. Jurisdiction case-law cites [3] when configured. No claim is left unsourced.

What you do get

One page per actor. Same format every time.

Single-page-per-actor by design. Half-inch margins, Calibri body, navy headings, red accent rule. The format is locked — what you see in the preview above is what the DOCX renders, paragraph-for-paragraph.

Forbidden-phrase enforcement

Blocklist scanned on every output.

The DOCX/PDF generators carry a FORBIDDEN_PHRASES tuple. A regression test scans every produced file for the phrases and fails the build if any appear. The memo never contains the literal words "depo­sition prep", "depo­sable", or similar.

Topology hash embedded

Same workbench result → same hash → same memo.

The manifest carries the upstream workbench topology hash. An opposing analyst running the same workbench result through the same engine version gets the same memo bytes. Reproducible by construction.