Purpose
Catch the gaps that get compliance documentation flagged in an exam or fail an internal QA review — missing elements, unsupported conclusions, stale or wrong regulatory citations, and weak audit trails — before the document leaves your desk. Produces a scored, line-referenced review with exact fixes so the author can revise in one pass.
When to Use
Invoke this skill when the user asks to:
- Review a SAR narrative, case write-up, or investigation memo before submission
- Check a Written Supervisory Procedure (WSP) or compliance procedure for completeness
- Verify regulatory citations in a compliance document are accurate and current
- QA a case file or documentation before it goes to a supervisor or examiner
- Assess whether a compliance write-up is audit-defensible
- Tighten the accuracy and clarity of a regulatory filing narrative
Inputs Required
Before starting, collect:
- Document type: SAR narrative / case memo / WSP / procedure / filing narrative / other
- Document text: The full text to review (paste or file reference)
- Regulatory framework: Applicable regime(s) — FINRA, SEC, BSA/AML, SAR (FinCEN), Reg BI, state, etc.
- Intended reader: Internal supervisor / QA / regulator/examiner / SAR filing
- Referenced facts or evidence: Underlying data the write-up is based on, if available
- Firm documentation standard: Any internal template or required fields, if one exists
If the regulatory framework isn't specified, infer it from the document type and content and state the assumption. If no internal standard is provided, apply general broker-dealer/fintech documentation norms.
Instructions
Step 1: Classify the Document and Set the Standard
Identify the document type and the review standard it must meet:
- SAR narrative: Must satisfy FinCEN's five essential elements — who, what, when, where, why (and "how" where applicable). Chronological, fact-based, self-contained.
- Case memo / investigation write-up: Must state the allegation/trigger, evidence reviewed, analysis, conclusion, and disposition with rationale.
- WSP / procedure: Must specify the who (responsible party/title), what (the control), when (frequency/trigger), how (the steps), and evidence of review/supervision.
- Filing narrative: Must be accurate, complete, internally consistent, and traceable to source data.
Note which standard applies and use it as the completeness rubric in Step 3.
Step 2: Factual Accuracy Check
Read every factual claim against the referenced evidence (or flag where evidence is absent):
- Flag any statement presented as fact that isn't supported by the provided data
- Flag internal inconsistencies (dates, dollar amounts, account numbers, entity names that don't reconcile)
- Flag conclusions that outrun the evidence (e.g., "customer knowingly structured deposits" without support)
- Flag hedged or ambiguous language where a definite fact is required, and vice versa
- Note any figure, date, or identifier that appears wrong or transposed
Step 3: Completeness Check
Against the standard from Step 1, list every required element and mark each: Present / Partial / Missing.
- For SAR narratives, confirm all five elements plus a clear statement of suspicious activity
- For case memos, confirm trigger, scope, evidence, analysis, conclusion, disposition
- For WSPs, confirm responsible party by title, control description, frequency, escalation path, and supervisory review
- Identify any required element that is missing entirely and any that is present but underdeveloped
Step 4: Regulatory Citation Integrity
Review every regulatory reference:
- Verify the citation format is correct (e.g., 31 CFR 1023.320, FINRA Rule 3110, SEC Rule 17a-4, 12 CFR / BSA references)
- Flag misattributed rules — a control cited to the wrong rule number or agency
- Flag stale references — rules that have been renumbered, superseded, or replaced (e.g., legacy NASD rules that migrated to FINRA rulebook)
- Flag missing citations — a stated obligation or control with no rule cited where one is expected
- Where a citation is wrong or missing, provide the likely correct reference and mark it as verify before filing — never present a corrected citation as confirmed without a directive to confirm against the source
Step 5: Documentation Standards & Audit Defensibility
Assess whether the document would hold up under examination:
- Traceability: Can each conclusion be traced to specific evidence?
- Objectivity: Is language neutral and fact-based, free of speculation, opinion, or emotional characterization?
- Self-containment: Can a reader who wasn't involved understand the case without outside context?
- Chronology & completeness of record: Are dates, actions, and decisions logged in order?
- Attribution & sign-off: Are preparer, reviewer, and dates present where required?
- Flag any passage that reads as editorializing, assumes undocumented knowledge, or leaves a decision unexplained
Step 6: Clarity & Language
Flag structural and language issues that weaken the record:
- Run-on or ambiguous sentences that obscure the fact pattern
- Undefined acronyms or internal jargon on first use
- Passive constructions that hide the actor ("the account was flagged" — by whom?)
- Redundancy or filler that dilutes the material facts
Step 7: Score and Prioritize
Assign a score of 1–5 to each of the six dimensions (Accuracy, Completeness, Citation Integrity, Documentation Standards/Defensibility, Clarity, plus an overall). Then produce a prioritized revision list ordered by severity: Blocker (would fail QA or exam) → Major (materially weakens the record) → Minor (polish). Each item references the specific location and gives the exact fix.
Output Format
- Structure: Summary scorecard table → dimension-by-dimension findings → prioritized revision list → citation verification table
- Length: 700–1,100 words
- Format: Structured document — scorecard as a table, findings as labeled sections with line-referenced bullets, revisions as a ranked list
- Sections: Scorecard (six dimensions, 1–5) → Factual Accuracy → Completeness (element checklist: Present/Partial/Missing) → Regulatory Citation Integrity (with verify-before-filing flags) → Documentation Standards & Defensibility → Clarity → Prioritized Revision List (Blocker/Major/Minor)
- Tone: Compliance QA reviewer — precise, neutral, evidence-first; no hedging, no softening of material issues
Quality Checklist
Before delivering output, verify:
- [ ] Every one of the six dimensions has a 1–5 score with at least one supporting finding
- [ ] The completeness check marks each required element explicitly as Present / Partial / Missing
- [ ] Every citation issue names the exact rule cited, why it's wrong or missing, and a suggested correction flagged "verify before filing"
- [ ] Each revision-list item is labeled Blocker / Major / Minor and points to a specific location in the document
- [ ] No suggested citation is presented as confirmed accurate without a directive to verify against the source
- [ ] Findings reference actual passages from the document — not generic advice
Notes
- For SAR narratives, the most common failures are missing "why it's suspicious" and conclusions unsupported by the transaction detail — check those first.
- Never treat a suggested citation correction as authoritative; regulatory rule numbers change, and this review flags them for human verification against the current rulebook or CFR.
- If evidence or source data is not provided, factual-accuracy findings are limited to internal consistency — say so explicitly rather than implying the facts were verified.
- A document scoring 5 on clarity but 2 on completeness still fails an exam; weight completeness and citation integrity highest when giving the overall score.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills