Purpose
Systematically scan reconciliation schedules to surface variances, unexplained balances, and stale reconciling items that a reviewer would otherwise miss under close-deadline pressure — delivering a ranked anomaly register with root-cause hypotheses and specific follow-up actions so the account can be signed off with confidence.
When to Use
Invoke this skill when the user asks to:
- Review balance sheet account reconciliations before close sign-off
- Investigate a variance between the GL balance and the supporting schedule
- Find stale or aging reconciling items that haven't cleared
- Check reconciliations for unusual movements or plug entries
- Prepare reconciliation packages for manager or auditor review
- Harden reconciliation controls across a set of accounts
Inputs Required
Before starting, collect:
- Account(s) under review: Account name, number, and type (cash, AR, AP, accruals, prepaids, fixed assets, intercompany, suspense, etc.)
- Reconciliation schedule: The current-period reconciliation detail — GL balance, sub-ledger/supporting balance, and listed reconciling items with dates and amounts
- Prior-period reconciliation: The same account's reconciliation from the prior close (for trend and stale-item comparison)
- Materiality threshold: The dollar or percentage threshold at which a variance requires investigation
- Expected activity: Any known drivers for the period (large invoices, one-off entries, period-specific accruals)
- Aging tolerance: How old a reconciling item can be before it's considered stale (e.g., 30/60/90 days)
- Close deadline context: Whether this is a pre-close review, hard-close review, or audit prep
If materiality or aging tolerance is not provided, default to the greater of $10,000 or 5% of the account balance for materiality, and 60 days for staleness — and state the assumption.
Instructions
Step 1: Confirm the Reconciliation Ties
Establish whether the reconciliation is arithmetically sound before hunting anomalies.
- Verify GL balance − supporting balance = sum of reconciling items (the reconciliation must foot)
- Flag any out-of-balance reconciliation immediately — this is a Priority 1 anomaly regardless of amount
- Confirm the supporting balance is sourced (sub-ledger, bank statement, amortization schedule, third-party confirmation) and not a manually keyed figure
- Note any reconciliation where the supporting balance equals the GL balance exactly with no independent source — this is a common sign of a "forced" reconciliation
Step 2: Variance Analysis
Compare the GL balance to prior period and to expectation.
- Calculate the period-over-period movement in absolute and percentage terms
- Flag movements exceeding materiality that are not explained by known expected activity
- Flag unexpected sign (e.g., a contra account with a debit balance, a liability with a debit balance, negative cash)
- Flag directional anomalies: accounts that should grow steadily (accumulated depreciation) but declined; accounts that should be near-zero (suspense, clearing) carrying a balance
- For each flagged variance, state the expected behavior, the observed behavior, and the gap
Step 3: Reconciling Item Scrutiny
Examine every line in the reconciling-items list.
- Stale items: Flag any item older than the aging tolerance; escalate items older than 90 days
- Unexplained items: Flag any reconciling item lacking a description, reference number, or owner
- Round-number / plug items: Flag suspiciously round amounts or items labeled "plug," "difference," "to balance," or "misc" — these often mask an unresolved error
- Recurring items: Flag items that appear unchanged across two or more consecutive periods — a true reconciling item should clear
- Offsetting items: Flag pairs of near-equal debit/credit reconciling items that may indicate a misposting rather than genuine timing differences
- Large single items: Flag any single reconciling item exceeding materiality
Step 4: Control and Timing Signals
Assess the reconciliation as a control artifact.
- Check whether preparer and reviewer are the same person (segregation-of-duties gap)
- Check the reconciliation date against the period-end date — reconciliations completed long after close reduce reliability
- Flag accounts with no reconciling items and no supporting detail where the balance is material (may be un-reconciled, not truly clean)
- Note any account that was materially different last period but is now zero without a documented clearing entry
Step 5: Build the Ranked Anomaly Register
Compile every flag into a single prioritized register. For each anomaly, capture:
- Anomaly ID and account
- Category: Out-of-balance / Variance / Stale item / Unexplained item / Plug / Control gap
- Amount at risk: Dollar exposure
- Priority: P1 (blocks sign-off) / P2 (investigate before close) / P3 (monitor)
- Root-cause hypotheses: 1–3 plausible explanations, most likely first
- Follow-up action: The specific next step and who should own it
Priority logic:
- P1: Reconciliation doesn't foot, forced/plugged balance, material unexplained variance, negative or wrong-sign balance
- P2: Stale item over aging tolerance, unexplained reconciling item over materiality, control gap on a material account
- P3: Sub-materiality stale items, minor recurring timing differences, documentation gaps
Step 6: Sign-Off Recommendation
State a clear recommendation: Ready for sign-off, Sign-off with noted follow-ups, or Do not sign — resolve P1 items first. Justify in two sentences.
Output Format
- Structure: Narrative diagnosis by step, followed by the anomaly register as a table, closing with the sign-off recommendation
- Length: 700–1,100 words
- Format: Prose for Steps 1–4 diagnosis; the anomaly register in Step 5 as a table (columns: ID, Account, Category, Amount, Priority, Root-Cause Hypotheses, Follow-Up Action); recommendation in bold
- Sections: Reconciliation Integrity → Variance Analysis → Reconciling Item Scrutiny → Control Signals → Anomaly Register (table) → Sign-Off Recommendation
- Tone: Accounting reviewer — precise, skeptical, evidence-driven; every flag tied to an amount or a specific rule, never a vague concern
Quality Checklist
Before delivering output, verify:
- [ ] The reconciliation footing check is performed and stated explicitly (does GL − support = reconciling items)
- [ ] Every anomaly in the register has a dollar amount, a priority, at least one root-cause hypothesis, and a follow-up action
- [ ] Priority assignments follow the P1/P2/P3 logic — out-of-balance and plugged items are always P1
- [ ] Stale items are measured against the aging tolerance (stated or defaulted), not flagged arbitrarily
- [ ] The materiality threshold used is stated, including any default assumption
- [ ] A single, unambiguous sign-off recommendation is given (no hedging)
Notes
- The most common hidden risk is the "forced reconciliation" — where the supporting balance is silently set equal to the GL to make the account foot; always test whether the support is independently sourced
- Recurring reconciling items that never clear are a red flag more often than large one-time items — persistence signals an unresolved error, not a timing difference
- Suspense, clearing, and intercompany accounts warrant the tightest scrutiny; a non-zero balance in a clearing account at period-end is almost always an anomaly
- This skill analyzes the reconciliation as presented; it does not re-perform the underlying accounting — recommend pulling source detail for any P1 item before adjusting
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