Explore the build library
consultance.ai
Deal Copilot
Build libraryControlsAboutCase studiesBlogBook an audit
consultance.ai

Systems

WorkflowsDeal CopilotPrivate knowledgeFinance automationBuild library

Company

About usCase studiesBlogContact

Start a conversation

Free AI auditfounder@consultance.ai

Teams

Investment bankingCFOs & financePrivate equityPrivate creditReal estateProperty managementProfessional servicesOther industriesAll industries

Legal

Privacy policyTerms of service

Our approach

Audit the workflow.Build in your environment.Hand over a system you own.
Build something useful © 2026 consultance.ai
Build library/Quarter End Finance Skills for Claude

Finance and data

Quarter End Finance Skills for Claude

For bank, insurer and oil and gas controllers: 14 prompts that run 22 open source finance skills in your own Claude to draft quarter end entries, with a tie out gate.

Medium setup4 steps14 prompts
Open the setup guide Discuss an implementation
Finance and data

Start with the guided setup

Follow the setup below in your own workspace. For a demonstration of the prompt-copying process, open the optional walkthrough.

Watch the setup walkthrough
Step 1 · setup
Three minutes, four steps, nothing to install by hand

Claude sets it up for you. You just paste.

Never used Claude? It is free and takes 30 seconds to open. Copy the instruction below, paste it into Claude, and it reads this page and walks you through everything, one question at a time.

  1. Tell Claude how to talk to you

    One tap. It changes how much Claude explains, and how slowly it goes. You can change it any time.

  2. Copy your setup instruction

    A short instruction plus a link to this page lands on your clipboard. First copy asks for your email once. That unlocks every button across the whole library.

  3. Open Claude in a new tab

    Free account, no card, 30 seconds. This tab stays open so you can come back.

    Open claude.ai
  4. Paste, send, and answer one question

    Claude reads this page, asks which computer you are on, then guides you step by step until it works. If anything errors, tell Claude what you see, and it fixes it with you.

▸Prefer the full prompt instead of the link? (optional)
Click to copy
I am comfortable copy-pasting and following instructions, but I am not a developer.
- Plain English. Define jargon the first time it appears.
- One step at a time, then wait for me to confirm before the next one.
- Tell me what success looks like at each step, and diagnose any error before moving on.

Follow the instructions below with those rules applied.

If you can browse the web, open and read this page in full first, it has the complete guide and every prompt you will run (the vault is under the-vault anchor): https://consultance.ai/library/quarter-end-finance-skills#the-vault . If you cannot open links, tell me and I will paste the page in, do not guess the prompts.

I want to set up the quarter end finance skills from github.com/GAJETOso/financeskills, an MIT licensed repo of 48 finance skills built by Ayotunde Oso. I will use the 20 quarter end jobs across five stages (Measure, Reconcile, Test, Value, Report), which take 22 skill folders, and then run the prompt vault on that page. Be my calm, practical guide from consultance.ai. One step at a time. Wait for my result after every step. Use Claude Opus 5.5; if I am on another model, tell me to select it in the model picker.

## Step one, ask me this first, nothing else
First message: "Which of these is you: (A) I have Claude Code or can install it, (B) I only use Claude in the browser or desktop app, or (C) I just want to look at one skill before installing anything?"

Wait for my answer. Then tell me which path and why. If my quarter end files are a folder of exports, recommend path A even if I picked B, and say why: Claude Code reads the files off disk and runs the skills' calculators.

## Words you will hear
- Skill: a folder with a SKILL.md instruction file, usually with a small Python calculator, that Claude loads when my request matches it.
- Claude Code: Claude working in a folder on my computer, able to read files and run commands.
- Terminal: the app where I type commands (Mac: Applications, Utilities, Terminal).
- Journal entry draft: the entry the skills prepare for me to post in my own ledger. Claude never posts it.

## The stage map (post name, then skill folder)
- Measure: Loan Loss Calculator = ecl-computation; Insurance Reserve Estimator = insurance-reserving; Well Shutdown Liability = aro-computation; Revenue Recognition Checker = revenue-recognition.
- Reconcile: Bank and Intercompany Reconciler = automated-reconciliation plus intercompany-accounting; Group Consolidator = corporate-consolidation; Cost Variance Investigator = variance-analysis plus budget-forecast; Crypto Asset Ledger = crypto-accounting. Which variance skill fits: variance-analysis is for manufacturing standard costs (material, labour, overhead); budget-forecast projects and compares scenarios but has no budget against actual function, so for banks and insurers prompt 08 works budget against actual by hand, with the working shown.
- Test: Audit Checklist Builder = audit-checklist; Fraud Investigator = forensic-accounting; Anomaly Detector = ai-anomaly-detection; Basel III Capital Checker = banking-compliance.
- Value: Oil and Gas Reserve Valuer = oil-gas-valuation; Cost of Capital Calculator = wacc-computation; Investment Case Tester = investment-analysis; Debt Restructuring Modeler = debt-restructuring.
- Report: Financial Statement Builder = statement-preparation; Ratio Analyst = financial-analysis; ESG Report Writer = esg-reporting; Earnings Call Reader = nlp-earnings-sentiment. Answers to the questions a board will ask come from vault prompt 13, which works from the statements these skills draft.
The repo has 48 skills: 43 ship in its Claude Code plugin and 5 are personal finance skills we do not need.

## Path A, Claude Code (this is a Terminal install)
1. `node --version`. Success: a version number. If "command not found", install the LTS version from nodejs.org, reopen Terminal, check again.
2. `mkdir quarter-end` then `cd quarter-end`. Put my quarter end exports in this folder later. Skills load only when I start Claude Code here.
3. Install the 22 skill folders, one line. The two flags at the end, `-a claude-code -y`, come from the installer's own README: install for Claude Code only, into this folder, with no questions asked.
`npx skills add GAJETOso/financeskills --skill ecl-computation insurance-reserving aro-computation revenue-recognition automated-reconciliation intercompany-accounting corporate-consolidation variance-analysis budget-forecast crypto-accounting audit-checklist forensic-accounting ai-anomaly-detection banking-compliance oil-gas-valuation wacc-computation investment-analysis debt-restructuring statement-preparation financial-analysis esg-reporting nlp-earnings-sentiment -a claude-code -y`
Success: it ends "Done!" and lists each skill as "✓ <name> (copied)" going to ./.claude/skills/<name>. If npx asks to install the skills package, say yes. To take all 48 instead: `npx skills add GAJETOso/financeskills -a claude-code -y`.
If I typed the line without the flags and the installer asked questions, pick Claude Code as the agent, Project as the scope, and Symlink. If `ls .claude/skills` then shows nothing, I picked Global: run the line again from inside the quarter-end folder with `-a claude-code -y` at the end.
4. `ls .claude/skills` must show the 22 names. The installer's first summary table may say ./.agents/skills; ignore it. The final ✓ list and this `ls` are the real check.
5. Optional proof the calculators run: `python3 .claude/skills/ecl-computation/scripts/calculate.py` prints a line starting "lifetime ECL". The calculators use only Python's standard library; nothing else to install.
6. `claude`, then `/model` and choose Opus 5.5. Ask "Which finance skills do you have?"
Bonus route, from the repo README and not tested by us: inside Claude Code, `/plugin marketplace add GAJETOso/financeskills` then `/plugin install finance-skills` (the plugin carries 43 skills, including all 22 above).

## Path B, the Claude app. Not a Terminal install.
1. On the GitHub repo page, click Code, then Download ZIP, and unzip it.
2. Open the skills folder and compress only the ecl-computation folder into a ZIP. Keep the folder name exactly ecl-computation.
3. In Claude: Settings > Capabilities, turn on Code execution and file creation.
4. Customize > Skills, click +, then + Create skill, then Upload a skill. Upload the ZIP. On Team or Enterprise, an owner may need to enable skills first.
5. Repeat for the other skills my stage needs, one at a time.
Do NOT tell me this is "not possible": uploading a skill ZIP is a normal Claude feature.

## Path C, no install
Open skills/ecl-computation/SKILL.md on GitHub, paste its text into this chat, and use it for the drill. Tell me plainly that the calculator does not run on this path, so every figure must be worked by hand and checked in vault prompt 12.

## Privacy
Nothing is hosted by consultance.ai. My files go only to my own Claude account. Before confidential ledgers, I use a Team or Enterprise plan, or a personal plan with model training turned off.

## First session drill
1. Open the vault on the page. Paste prompt 01 and answer its questions one block at a time.
2. Paste prompt 02, the calibration run. Good output: Part 1 provisions 2,000, 3,000, 3,000, 3,750 and 7,500, total 19,250, draft entry 3,250. Part 2 stops on loan L5 because its LGD is missing, and names the loan tape. After I reply "for this test, exclude L5": total 71,800 and a draft charge of 15,800.
3. If any line says MISMATCH, stop and fix it before real data. If all match, run prompt 03 with my own trial balance and my view of the quarter.
4. Before I trust any real output, run prompt 12, the tie out. It says TIES, BLOCKED with both figures, or OPEN with the missing documents. Every entry stays a draft for me to post.
5. Then offer the prompt that fits my entity type from prompt 01, one at a time.

This drafts work for me to review and post. It does not sign the accounts, and my auditor still reviews the result.
Step 2 · run it on your data

Step 1 set it up. These 14 prompts do the work.

the vault

The 14 prompts

Grab the whole pack as one file, or tap any prompt below to copy it on its own. Placeholders that look like {{THIS}} get swapped for your own numbers — and if you ran Step 1, Claude fills them in for you.

One .md file · all 14 prompts, numbered, in order · nothing left out.
Click to copy
<role>A colleague who has closed many quarters as a controller at banks, insurers and oil and gas producers. Neutral: you draft and check and expect most of what you see to be normal. You run the skills on my files; I review, post and sign.</role>

<surface>
- One schedule or a few files I can upload: a private Claude Project is correct. Say so.
- A folder of exports, data that must stay on a managed machine, or a repeat run: Claude Code, which reads files off disk and runs the calculators.
STOP AND REROUTE, do not continue, when:
1. I paste a file path, folder or file listing instead of content: say this is a Claude Code job and stop.
2. The material is too large or truncated: list what is missing and compute nothing from the part you saw.
3. I am running the same pass over many entities by hand. Recommend Claude Code and stop.
4. A figure needs a document I have not provided: ask once, naming it, then stop on that figure.
</surface>

<model>Claude Opus 5.5 for every prompt; select it in the model picker (Claude Code: /model). Claude Sonnet 5 only for a first read of a very large export set, never for judgement or prompt 12. Do not switch model inside a prompt.</model>

<privacy>My files go only to my own Claude account, never to consultance.ai. Confidential ledgers: Team or Enterprise plan, or model training off.</privacy>

<onboarding>
Ask one block at a time and wait for each answer.
1. TYPE, each option names its prompts.
   (A) Bank or lender: 04 loan loss allowance, 09 capital check, then 08, 11 to 13.
   (B) Property and casualty insurer: 05 IBNR, then 08, 11 to 13.
   (C) Oil and gas producer: 06 retirement obligations, 10 reserve valuation, then 08, 11 to 13.
   (D) Other company: 07 revenue, then 08, 11 to 13. Any type can run the trade receivables matrix in 04.
   Life insurance, pensions and IFRS 17 measurement models are not covered: say so and stop.
2. STAGE. Measure (04 to 07), Reconcile (08), Test (09), Value (10), Report (11, 13), or all. Prompt 12 always runs before the signer sees anything.
3. MODE. CONVERSATION (default): I state my view of the quarter, you ask for what you need and run only the checks that bear on it. FULL SEQUENCE: every prompt for my type, in order.
4. DATA SOURCE. (A) files in this private Project; (B) figures pasted in chat; (C) Claude Code on a folder; (D) a governed ledger connection my firm already runs (this pack sets none up); (E) a mix.
5. TOKENS. {{ENTITY}}, {{PERIOD_END}}, {{FRAMEWORK}} (IFRS or US GAAP), {{CURRENCY}}, {{MATERIALITY}} (below this a point gets no tag), {{TOLERANCE}} (tie tolerance, default 1 unit), {{SIGNER}} (the named person who signs).
</onboarding>

<skill_check>
Before a stage runs, confirm its skills are loaded (Claude Code: ls .claude/skills; Claude app: Customize, Skills). If one is missing, name it, send me to the setup guide, and stop. A pasted SKILL.md with no install cannot run its calculator: show every step so prompt 12 can redo it.
Each prompt heading names its skills; each prompt's skill notes say what its calculators really return. None posts to a ledger or knows my policies.
</skill_check>

<evidence_tiers>
TIER 1: primary record (ledger, subledger, loan tape, triangle, engineer's estimate, signed contract, policy).
TIER 2: a figure keyed or exported by a person (workpaper, summary, a figure I type).
TIER 3: commentary, decks, emails. TIER 3 raises a question, never supplies a number.
Every load bearing figure carries tier and source, e.g. [T1 loan-tape.csv row B].
</evidence_tiers>

<rules>
- Never invent an input. A missing rate, balance, factor or policy is OPEN: ask once naming the document; do not estimate it.
- Two sources disagree on a figure: show both and stop on that figure. Never average or pick the tidier one.
- Full precision, round at the end. A tie beyond {{TOLERANCE}} blocks; a difference from display rounding alone is not a break.
- Show the working (formula, inputs, result) so a person can redo it by hand.
- You draft each journal entry ready for me to post; never say it is posted.
FLAGS
- Big picture first: prompt 03 records what moved and my view before any check.
- Test every point against the prompt's <normal> list (pattern | why normal | abnormal if) and check the abnormal condition it names.
- Tag each point: CHANGES THE DECISION (moves a figure by at least {{MATERIALITY}} or changes whether the entry should be signed), WORTH A QUESTION (the answer could move a figure by at least {{MATERIALITY}}), EXPLAINED BY CONTEXT (name the context). Below {{MATERIALITY}}, conventions, rounding and timing the documents resolve: one untagged line, Checked, normal.
- Missing optional data is OPEN, never a flag. A point closed by context or by me stays closed.
- Arithmetic and recurrence outrank the normal list: a one time item that recurs every quarter is not one time.
- Never explain away an arithmetic FAIL. That rule is for arithmetic only, never for judgement points.
- A block stops the verdict, not the analysis. BLOCK PROTOCOL for every review gate: on a break print BLOCKED with both figures, keep running the other checks, produce no signed off version, print the default if nobody acts (every JE stays a draft, nothing is posted), and name who unblocks it ({{SIGNER}}, fixing the input or entry).
- Every working prompt ends with its tagged points and one Checked, normal line.
</rules>

<adapt>Edit {{MATERIALITY}} and {{TOLERANCE}} per entity (never widen tolerance after a break); add your own normal patterns, same form; your document names in prompt 03.</adapt>

Confirm my answers in five lines, then name the next prompt. On a new install, prompt 02 first.
Click to copy
<role>Same colleague, testing the pack on a synthetic sample with a known answer.</role>

<task>
Run ecl-computation on this sample (Claude Code: call provision_matrix and ecl_single_period and show the call). IFRS 9, USD, {{MATERIALITY}} 2,000, {{TOLERANCE}} 1, discount factor 1.0.
PART 1, trade receivables provision matrix. Opening allowance 16,000, no write offs. Current 400,000 at 0.5%; 1 to 30 days 150,000 at 2%; 31 to 60 days 60,000 at 5%; 61 to 90 days 25,000 at 15%; over 90 days 15,000 at 50%.
PART 2, loan tape, general approach. Opening allowance 60,000 at the start of the quarter; since then write offs of 5,000 and recoveries of 1,000 have been posted to the allowance. Policy: stage 2 if more than 30 days past due, on the watchlist, or lifetime PD at least double origination; stage 3 if more than 90 days past due.
L1 stage 1, 0 dpd, EAD 1,000,000, 12m PD 1.0%, lifetime PD 5% (origination 4%), LGD 40%.
L2 stage 2, 35 dpd, EAD 500,000, lifetime PD 8%, LGD 40%.
L3 stage 3, 110 dpd, EAD 200,000, PD 100%, LGD 25%, property appraised last month.
L4 stage 1, 0 dpd, amortising, EAD 300,000 (340,000 last quarter), 12m PD 1.5%, LGD 40%.
L5 stage 2, 40 dpd, EAD 150,000, lifetime PD 10%, LGD not provided.
</task>

<expected_output>
Expected output columns per part: line, inputs, working, result, tier. MATCH is judged on substance, not layout.
PART 1: 2,000; 3,000; 3,000; 3,750; 7,500 (balance times rate). Total 19,250 on 650,000, coverage 2.96%. Charge 19,250 less 16,000 = 3,250. Draft JE: Dr impairment loss on trade receivables 3,250, Cr loss allowance 3,250.
PART 2 first pass: L1 0.01 x 0.40 x 1,000,000 = 4,000. L2 0.08 x 0.40 x 500,000 = 16,000. L3 1.00 x 0.25 x 200,000 = 50,000. L4 0.015 x 0.40 x 300,000 = 1,800. L5 OPEN, LGD missing from the loan tape. Total, charge and JE withheld; the run says it stopped, names the loan tape and L5, and does not estimate an LGD.
Checked, normal (no tag): L4 falls because it amortises; L3 at 25% of exposure because of collateral; stage 1 coverage under 1%; the current receivables bucket carries a rate.
PART 2 second pass, after I reply "for this test, exclude L5": total 71,800 (stage 1 5,800; stage 2 16,000; stage 3 50,000); coverage 3.59%. Ledger before entry 60,000 less 5,000 plus 1,000 = 56,000; charge 71,800 less 56,000 = 15,800. Draft JE: Dr credit impairment charge 15,800, Cr loan loss allowance 15,800.
</expected_output>

<review_gate>Print MATCH or MISMATCH per line. If a figure differs, the L5 stop did not fire, or a normal item was tagged CHANGES THE DECISION or WORTH A QUESTION, name the line and STOP; load no real data until it matches. If all match, recommend prompt 03.</review_gate>
Click to copy
<role>Same colleague, reading the quarter before testing anything.</role>

<task>
Use the data source from prompt 01. One page on {{ENTITY}} for the quarter ended {{PERIOD_END}}:
1. What the entity is and how it earns money, three lines.
2. My view in my own words: what I expect the allowance, reserves, obligations or revenue to do and why. Ask if I have not said. Quote it back unchanged.
3. What moved the estimates (rates, prices, claims, failures, new lending or wells); news is TIER 3 context.
4. Inputs checklist for my TYPE, each HAVE or OPEN with its tier.
   All: trial balance at {{PERIOD_END}}; prior quarter ledger closing; each estimate's policy; last quarter's workpaper.
   Bank: loan tape (stage, days past due, PD, LGD, EAD, undrawn), staging policy, scenario weights, overlays, write offs and recoveries.
   Insurer: paid and incurred triangles, gross and ceded; case reserves; last actuarial range; expected loss ratios for immature years.
   Oil and gas: ARO layers with rates; engineer's estimates; wells drilled, sold, plugged; reserve report; price series.
   Revenue: contracts and modifications; SSP support; billing and delivery schedules.
   Reconcile, Test, Report: bank statements, intercompany balances, journal entry listing, prior statements.
5. Which prompts to run, in order, for my view.
</task>

<trap>Last quarter's workpaper closing is not the opening balance. Take the opening from the ledger; a difference between the two is a finding before any estimate runs.</trap>

<stop>If the trial balance at {{PERIOD_END}} or the ledger opening for an estimate is not provided, name the document and stop; do not start an estimate from a workpaper alone.</stop>

<review_gate>Show the picture and my view. Run no checks until I confirm both.</review_gate>
Click to copy
<role>Same colleague, as credit risk accountant for the quarter.</role>

<task>
Use the data source from prompt 01; use ecl-computation (provision_matrix, ecl_single_period, lifetime_ecl, scenario_weighted_ecl). Under CECL every loan carries lifetime loss and there are no stages; skip step 2.
1. Tie EAD first: tape EAD equals gross loans in the trial balance plus undrawn commitments at the policy conversion factor; list reconciling items.
2. Staging: apply every policy trigger to every loan (days past due backstop, watchlist, forbearance, lifetime PD against origination), from the raw fields, never the file's stage column. List loans whose stage differs from the policy result, with ECL both ways.
3. ECL per loan or segment with the skill, working shown. Show each scenario, its weight, and the weighted result; weights must sum to 1.
4. Overlays listed separately (owner, amount, stage, release condition), never netted into model ECL.
5. Roll forward: ledger opening, charge, write offs, recoveries, closing (= step 3 plus overlays); stage transfer table (count, EAD).
6. Draft JE and coverage by stage. Banks: carry the closing allowance to prompt 09.
Trade receivables, any TYPE: simplified approach with provision_matrix, same roll forward and JE.
</task>

<normal>
- Allowance rises while loans fall | stage 2 migration or a worse forecast | no transfers and no forecast change.
- Stage 1 coverage well under 1% | 12 month loss on performing loans | above stage 2 coverage.
- Stage 3 loan with ECL well below exposure | collateral drives LGD | appraisal stale, or no costs to sell.
- Allowance falls after write offs | write offs use provision already held | charge also falls while arrears rise.
- Amortising balance falls each quarter | scheduled repayment | no cash received.
- Stage 2 loan at 0 days past due | PD or watchlist trigger | no trigger recorded.
- Current receivables bucket carries a rate | lifetime losses on all trade receivables | rate zero, or above older buckets.
</normal>

<trap>A loan past the days past due backstop or on the watchlist but left in stage 1 carries 12 month ECL instead of lifetime. The book total still looks plausible; the error is several times that loan's ECL. A stage kept against a trigger needs a documented rebuttal; without one the loan moves.</trap>

<stop>If the staging policy, PD, LGD or EAD for any loan, or the opening allowance per the ledger is not provided, name the document and loan, mark it OPEN, and do not estimate it. Totals, roll forward and JE are withheld until it is supplied or I exclude that loan by name.</stop>

<review_gate>Roll forward closing must equal step 3 plus overlays, and the ledger allowance after the draft JE must equal that closing, within {{TOLERANCE}}. Otherwise apply the BLOCK PROTOCOL and do not produce a signed off version.</review_gate>
Click to copy
<role>Same colleague, beside the reserving actuary. You draft the accounting; the actuary owns the estimate.</role>

<task>
Use the data source from prompt 01; use insurance-reserving (development_factors, project_ultimates; its ibnr() is total unpaid, not IBNR).
1. Triangles paid and incurred, gross and ceded, kept separate; large and catastrophe losses alone. Latest diagonal ties to ledger paid and case reserves.
2. Actual against expected emergence since last quarter.
3. Age to age factors, the selected factor and why (all year, latest three, a named outlier excluded), and the tail factor with its source.
4. Ultimates by chain ladder; for the two latest accident years also Bornhuetter Ferguson on my expected loss ratio, and which you would weight.
5. IBNR = ultimate less paid less case, per accident year and total.
6. Roll forward: ledger opening, paid, current year incurred, prior year development by accident year, closing; gross, ceded, net.
7. Booked reserve against the actuary's central estimate and range. Draft JE for the IBNR movement.
</task>

<normal>
- Latest year has large IBNR and little paid | claims report and settle over years | short tail line still large at 24 months.
- Incurred factor below 1.00 | salvage, subrogation, case takedowns | persistent at every age on a line with no recoveries.
- Paid and incurred ultimates differ on recent years | paid is slower and more leveraged | gap widens one way every quarter.
- Favourable prior year development | prudent margin released as years mature | release exactly fills an earnings gap.
- Booked above central estimate | margin inside the range | outside the range, or margin moves with no memo.
</normal>

<trap>Chain ladder on the latest accident year multiplies a small, noisy paid number by large cumulative factors, so one early claim swings the ultimate. For immature years anchor on Bornhuetter Ferguson with an expected loss ratio from pricing or plan, never from the chain ladder output.</trap>

<stop>If a triangle cell is missing, the latest diagonal does not agree to the ledger, or an immature year has no expected loss ratio, name it, do not fill it, and stop on every year that uses it, directly or through a factor; other years still run.</stop>

<review_gate>Paid plus case plus IBNR must equal ultimate per year, and the roll forward must tie to the ledger within {{TOLERANCE}}. On a break apply the BLOCK PROTOCOL (the actuary joins {{SIGNER}} to unblock).</review_gate>
Click to copy
<role>Same colleague, owning the asset retirement obligation schedule.</role>

<task>
Use the data source from prompt 01; use aro-computation (initial_aro, accretion_schedule, revision_adjustment).
1. Every well and facility, shut in and inactive included, carries an obligation.
2. Layers by well: undiscounted cost, inflation, settlement year (matching the reserve report's economic limit), rate when recognised.
3. Accretion on opening layers first, each at its own rate. US GAAP: operating expense. IFRS: unwinding is a finance cost.
4. New wells: initial ARO at the current (US GAAP: credit adjusted risk free) rate.
5. Revisions. US GAAP: upward = new layer at today's rate; downward at the rate of the layers it reduces. IAS 37 and IFRIC 1: whole provision at the current rate each period. Say which applies.
6. Settlements: actual cost against carrying amount, gain or loss; current and non current split.
7. Roll forward: ledger opening, incurred, accretion, revisions, settlements, acquisitions and disposals, closing. Draft JEs.
</task>

<normal>
- ARO rises with no new wells | accretion | rise exceeds opening times rate with no revision.
- Downward revision does not reverse to the original amount | old layer rate (US GAAP) | discounted at today's rate.
- US GAAP ARO flat when rates fall | layers are not remeasured for rate | all layers restated at today's rate.
- IFRS provision jumps when rates fall | full remeasurement | jump with no rate or cost change.
- Loss on settlement | old wells cost more to plug | repeated losses with no upward revision on the rest.
</normal>

<trap>Discounting every layer at this quarter's rate looks consistent and is wrong under US GAAP: existing layers keep their original credit adjusted rates; only new wells and upward revisions use today's rate. When rates have risen, the single rate approach understates the liability.</trap>

<stop>If the rate for any layer, the settlement year, or the engineer's estimate behind a revision is not provided, name the well and document and stop on that layer; do not assume today's rate for an old layer.</stop>

<review_gate>Roll forward closing must equal the sum of layers and the ledger within {{TOLERANCE}}. On a break apply the BLOCK PROTOCOL.</review_gate>
Click to copy
<role>Same colleague, reviewing revenue for the quarter.</role>

<task>
Use the data source from prompt 01; use revenue-recognition (allocate_transaction_price, over_time_revenue, significant_financing_split, contract_liability).
1. New contracts and modifications; each modification a separate contract, a prospective change, or a cumulative catch up.
2. Performance obligations and transaction price, with variable consideration and the constraint reassessed at {{PERIOD_END}}.
3. Allocate on SSP with the skill, naming the evidence per obligation: observable, adjusted market, cost plus margin, or residual (only when price is highly variable or uncertain).
4. Timing: over time (cost to cost progress, uninstalled materials and wasted costs left out of costs incurred) or point in time (control evidence, cut off in the last days of the quarter).
5. Financing component where payment and transfer are over a year apart.
6. Contract balance roll forward (opening, billings, revenue from opening liability and from current billings, catch ups, closing); contract assets separate, netted only within a contract. Draft JEs.
</task>

<skill_notes>
- allocate_transaction_price splits in proportion to the SSPs you pass; it cannot tell list price from SSP.
- over_time_revenue is straight line by months elapsed, to date. Use it for stand ready obligations such as support; for cost to cost progress compute costs incurred over total estimated costs by hand. This quarter is the to date figure less revenue recognised before the quarter.
- contract_liability returns zero when revenue exceeds billings; that excess is a contract asset, recorded separately.
</skill_notes>

<normal>
- Deferred revenue rises while revenue falls | annual billing in a renewal quarter | billings flat, or no invoices behind it.
- Contract asset appears | milestone billing lags progress | ages past the milestone date.
- Revenue with no invoice | over time ahead of billing | no contract asset behind it.
- Late quarter spike | buying seasons | bill and hold, unusual terms, or credit notes early next quarter.
</normal>

<trap>Allocating on list price looks like SSP and is not when deals are routinely discounted: it shifts revenue to the obligations delivered first. Use observable standalone sales and discount bands before allocating.</trap>

<stop>If the signed contract or modification, SSP evidence for an obligation, or the billing schedule is not provided, name the contract and document and stop on that contract; do not default to list price. Other contracts still run.</stop>

<review_gate>Allocated prices must sum to the transaction price and the roll forward must tie to the ledger within {{TOLERANCE}}. On a break apply the BLOCK PROTOCOL.</review_gate>
Click to copy
<role>Same colleague, making the books agree.</role>

<task>
Use the data source from prompt 01. Run the parts that apply; each part stands alone.
1. Bank (automated-reconciliation). Adjusted bank = statement plus deposits in transit less outstanding payments; adjusted book = ledger plus bank items not yet booked. Every reconciling item singly with date, amount and age; draft JEs for unbooked fees and interest.
2. Intercompany (intercompany-accounting). Agree each pair; assign every difference to one side with its document; eliminate intercompany sales and unrealised profit in inventory still held.
3. Group (corporate-consolidation) (untested). Translate assets and liabilities at closing rate, income at average, equity at historical. Goodwill and NCI per {{FRAMEWORK}} (US GAAP: NCI at fair value; IFRS: elected per acquisition); NCI roll forward.
4. Cost variances (untested). Manufacturers (variance-analysis): material, labour and fixed overhead variances. Banks, insurers, service companies: budget against actual by line, split into volume, price, mix and timing against the phased budget, by hand with working shown. Never force standard costing onto a bank.
5. Crypto (crypto-accounting) (untested), if held: US GAAP fair value through net income at the principal market cut off price (ASU 2023-08); IFRS cost less impairment (IAS 38); units agreed to the custodian.
</task>

<skill_notes>
- exact_match pairs only when amount AND reference are identical; normalise references first or match on amount and date, and list what is left.
- tolerance_match gives each item its closest candidate in order, so one item can take another's match; check date and payee. one_to_many_match returns the FIRST combination of up to four items that sums to the target: a candidate until the remittance names the items.
- match_balances returns receivable less payable; positive means the receivable is higher.
- consolidated_equity_check ignores NCI and translation reserve: a reasonableness figure, not a tie.
- variance-analysis needs standard quantity and hours for ACTUAL output (flexed), not budget; positive is favourable.
- budget-forecast has no budget against actual function (scenario_delta only).
- realized_gain_fifo is FIFO only. cost_less_impairment impairs to the lowest price seen in the period and never reverses; under IFRS pass the recoverable amount at the test date instead.
</skill_notes>

<normal>
- Outstanding payments | presented next month | over 90 days, round, or voided and reissued.
- Deposits in transit | received after bank cut off | not cleared early next statement.
- Intercompany difference at close | goods or cash in transit | same difference across quarters, or no document.
- Seasonal line misses budget | budget phased in equal months | year to date also misses.
- Unfavourable price variance with favourable usage | better grade of material | both unfavourable with no volume change.
</normal>

<trap>A reconciliation that lands on zero by netting reconciling items into one line can hide two offsetting errors. List every item singly; a residual that is exactly offset by an unexplained item is a break, not a tie.</trap>

<stop>If the bank statement, either side of an intercompany pair, the standard cost card or phased budget, the exchange rates or the custodian confirmation is not provided, name the document and stop on that part; the other parts still run. Do not match or eliminate on a guess.</stop>

<output_format>Per part: inputs with tier, working, result, draft JEs; every reconciling item on its own line.</output_format>
<review_gate>Adjusted bank must equal adjusted book, and eliminations must net to zero, within {{TOLERANCE}}. An unexplained difference triggers the BLOCK PROTOCOL.</review_gate>
Click to copy
<role>Same colleague, asking the auditor's questions first. Most entries are fine.</role>

<task>
Use the data source from prompt 01.
1. Population first: opening trial balance plus the quarter's journal entry listing must equal the closing trial balance, account by account.
2. Fraud Investigator and Anomaly Detector (forensic-accounting, ai-anomaly-detection): entries after close, rare users, seldom used accounts, manual top side entries, preparer equal to approver, no approver, amounts just under an approval limit; round number ratio against the entity's own prior quarters.
3. (untested) Benford only for about 1,000 or more amounts across several orders of magnitude with no floor or cap; otherwise say it is not valid.
4. Outliers by account against the same quarter last year.
5. Audit Checklist Builder (audit-checklist): request list with owner, due date and trial balance tie.
6. Banks, Basel III Capital Checker (banking-compliance; tested on CET1 only, the rest untested): CET1, tier 1, total capital, leverage, LCR against the minimum plus every buffer that applies, with the regulator's risk weights and prompt 04's allowance.
</task>

<skill_notes>
- zscore_anomalies uses the whole series' standard deviation, outlier included: with 10 values or fewer it can never pass 3 and returns nothing. Use iqr_anomalies on short series and say so.
- rolling_deviation_anomalies tests nothing until it has more than 12 points.
- benford_test's chi square flag fires on most real ledgers of thousands of entries; read which digits deviate, not the flag.
- buffer_check defaults to 4.5% plus 2.5% only. Pass every buffer that applies; a US stress capital buffer replaces the 2.5%.
</skill_notes>

<normal>
- Quarter end spike in manual entries | accruals, reclasses, consolidation | after close, unapproved, or reversed early next quarter.
- Round accruals, rent and bonus | estimates and contracts are round | round amounts in revenue, or posted by an unusual user.
- Controller posts and a second person approves | normal segregation | the same user prepares and approves.
- Capital ratio falls after a provision | allowance reduces CET1 | falls below minimum plus buffers.
</normal>

<trap>Testing entries before proving the listing is complete tests whatever was exported; a clean result on an incomplete population says nothing. Roll the listing to the trial balance first, and name any account that does not roll.</trap>

<stop>If the listing does not roll to the trial balance, or a bank's risk weights, capital definitions or applicable buffers are not provided, name the gap and stop on that test; report the other tests separately.</stop>

<output_format>Population roll by account; exceptions (entry, date, user, account, amount, characteristic, tag); Benford or why skipped; outliers; request list; capital table (required, actual, surplus).</output_format>
<review_gate>If the population does not roll within {{TOLERANCE}}, print BLOCKED and do not call the testing clean. Default: testing incomplete. Unblocks: {{SIGNER}} supplying the full listing.</review_gate>
Click to copy
<role>Same colleague, checking valuations behind impairment and disclosure.</role>

<task>
Use the data source from prompt 01. Run the parts that apply; each stands alone.
1. Oil and Gas Reserve Valuer (oil-gas-valuation). SEC measure: 12 month first day average price, property differentials, flat costs, 10% discount; standardised measure after tax, PV-10 before. The full cost ceiling excludes accrued ARO outflows. Successful efforts impairment (ASC 360) uses the entity's own forward prices. State which applies.
2. Cost of Capital Calculator (wacc-computation) (untested): cost of equity, beta relevered at target structure, after tax cost of debt; IAS 36 value in use needs a pre tax rate.
3. Investment Case Tester (investment-analysis) (untested): NPV, IRR, payback on the cash flows provided.
4. Debt Restructuring Modeler (debt-restructuring) (untested): waterfall by seniority, recovery, haircut, 10% test of new against old terms.
</task>

<skill_notes>
- pv10 discounts volume times one netback, end of year from year 1, with no future development capex, abandonment or tax; subtract those by year before calling it PV-10.
- decline_production's first year is the initial rate with no decline.
- npv puts the first cash flow at time 0; Excel's NPV discounts it a period. Do not mix them.
- irr needs a sign change; with more than one, report NPV, not IRR.
- npv_of_terms values interest only with a bullet; discount amortising terms by hand. The 10% test discounts the new cash flows, fees included, at the ORIGINAL effective rate, not today's.
- waterfall pays strictly in the order given; split pari passu claims pro rata by hand.
</skill_notes>

<normal>
- Standardised measure falls, production flat | the 12 month average fell | falls while the average rose.
- Ceiling write down after spot recovered | trailing average lags spot | average and cushion both flat.
- Undeveloped reserves drop out | five year development rule | no plan change recorded.
- IRR above cost of capital, NPV negative | several sign changes in cash flows | a single sign change.
</normal>

<trap>Using spot or management's forecast deck in the SEC standardised measure looks current and is wrong: the rule fixes the 12 month first day average and a 10% rate. Keep the disclosure measure and the internal valuation separate.</trap>

<stop>If the reserve report, any month of the 12 month price series, the development and abandonment costs, the original effective rate, or the cash flows for a case are not provided, name the document and stop on that valuation; do not average 11 months or substitute spot. The other parts still run.</stop>

<output_format>Price; valuation by year (volume, netback, costs, discount factor, PV); ceiling or impairment test; WACC; investment table; waterfall; 10% test.</output_format>
<review_gate>Discounted values must recompute from the cash flows and rate within {{TOLERANCE}}. On a break apply the BLOCK PROTOCOL; no impairment entry is drafted.</review_gate>
Click to copy
<role>Same colleague, turning the close into draft statements.</role>

<task>
Use the data source from prompt 01, only with figures that passed prompts 04 to 10.
1. Financial Statement Builder (statement-preparation): balance sheet, income statement, indirect cash flow from balance sheet movements; opening cash plus net change equals closing with no balancing line.
2. Ratio Analyst (financial-analysis): liquidity, profitability, DuPont, solvency, each defined, against last quarter.
3. ESG Report Writer (esg-reporting) (untested), if in scope: framework named (IFRS S1 and S2, or ESRS), boundary as the financial statements, method and source per metric.
4. Earnings Call Reader (nlp-earnings-sentiment) (untested), with a transcript: remarks and questions read separately; hedging against last quarter.
</task>

<skill_notes>
- operating_cash_flow_indirect takes only receivables, inventory and payables movements (closing less opening). Every other working capital line goes in other_noncash, signed and listed; leaving one out is how a plug appears.
- financial-analysis returns no ROE, DuPont or debt to equity on nil or negative equity; say not meaningful.
- esg-reporting and nlp-earnings-sentiment have no calculator; every figure needs a source line, or it is a question.
</skill_notes>

<normal>
- Operating cash flow below net income | receivables growth or prepayments | persists with no sales growth.
- Cautious forward looking language | boilerplate | replaces confident language, or comes with a guidance cut.
- Scope 3 above scope 1 and 2 | value chain heavy business | method changed with no restatement.
</normal>

<trap>A cash flow statement that ties only through a balancing line hides a misclassified or missing item. Every line comes from a balance sheet movement or a named non cash item, and the balancing line is zero.</trap>

<stop>If a figure from prompts 04 to 10 is BLOCKED or OPEN, name it and do not place it in the statements; show that line as OPEN.</stop>

<review_gate>Balance sheet must balance and cash must prove within {{TOLERANCE}}. On a break apply the BLOCK PROTOCOL; statements stay drafts.</review_gate>
Click to copy
<role>An independent second reader. You did not prepare these figures; you recompute them.</role>

<task>
Recompute every load bearing figure from this run a second way from the TIER 1 inputs, not from earlier output.
1. Recompute each estimate (Claude Code: the skill's calculate.py; chat: by hand, each step shown).
2. Roll forward per estimate: ledger opening plus charge less write offs or settlements plus recoveries equals closing. Recompute the charge yourself; a draft formula can omit a component and look tidy.
3. Ledger: balance before the draft JE plus the JE equals the closing estimate.
4. Subledger totals equal the ledger control accounts.
5. Cross prompt: the charge in the draft statements equals the JE.
6. Tiers: any TIER 3 figure used as a number is a FAIL.
7. Carry forward only CHANGES THE DECISION points, each with corrected figure and JE.
</task>

<trap>The same calculation on the same inputs agrees with itself and proves nothing. Recompute from source records by a different route; tie to the ledger, not the workpaper.</trap>

<stop>Any difference beyond {{TOLERANCE}} in steps 1 to 5 is a FAIL: show both figures, name the line, do not explain or round it away, and do not reach a verdict. Keep running the other ties so every break is listed.</stop>

<output_format>
Tie table: figure, first result, second result, difference, PASS or FAIL.
Verdict: TIES (ready for {{SIGNER}}), BLOCKED (arithmetic break, listed) or OPEN (inputs missing, listed).
Corrected figures and JEs for each CHANGES THE DECISION point.
Default if nobody acts: every JE stays a draft and nothing is posted.
</output_format>
<review_gate>Only {{SIGNER}} clears a BLOCKED line, by fixing the input or entry; then this prompt runs again. A signed reason does not clear an arithmetic break.</review_gate>
Click to copy
<role>Same colleague, preparing the finance director for the board.</role>

<task>
Use the data source from prompt 01, only with figures prompt 12 marked PASS.
1. The eight to ten questions a board or audit committee is most likely to ask, ranked by the movement behind each: largest estimate moves, CHANGES THE DECISION points, capital or covenant headroom, anything OPEN.
2. Answer each in two or three sentences citing figure, tier, source and the roll forward line behind it; say an OPEN or BLOCKED figure plainly, with what closes it and who owns it.
</task>

<trap>An answer written to reassure fails the follow up. State the number and what drove it, including what went the wrong way; "prudence" or "market conditions" must name the roll forward line or say it is not yet known.</trap>

<stop>If prompt 12 has not run, or a figure a question needs is BLOCKED or OPEN, write the question and "not yet answerable" with the missing item named; no number.</stop>

<output_format>Table: question, draft answer, figure and tier, roll forward line, owner. Then questions not yet answerable.</output_format>
<review_gate>Every figure in an answer must appear in prompt 12's tie table as PASS; if one does not, do not produce the table. The finance director edits and owns every answer.</review_gate>
Click to copy
<role>Same colleague, producing a view of this run.</role>

<task>
From prompt 12's output in this conversation, write one self contained HTML file, quarter-end-view.html, with no external scripts, fonts or network calls (Claude Code: write it and print the path).
Sections: verdict banner with the default if nobody acts; tie table; roll forwards; CHANGES THE DECISION points with corrected figures; Checked, normal line; OPEN inputs by document.
</task>

<trap>A figure restated from memory drifts from the tie table.</trap>

<stop>If prompt 12 has not run in this conversation, do not produce the view; say to run prompt 12 first.</stop>

<review_gate>Take every figure from prompt 12's output; if one is not there, do not produce the file.</review_gate>
Source repo
https://github.com/GAJETOso/financeskills

The code is public and free. The setup instruction above installs and wires it for you. You never need to open this link.

From a working build to an operating system.

For bank, insurer and oil and gas controllers: 14 prompts that run 22 open source finance skills in your own Claude to draft quarter end entries, with a tie out gate.

Path A · free

Run the build independently

The setup rail and every prompt above are free and stay free. The cost is your time, and the risk of wiring it wrong on live data.

Back to the prompts
Path B · done with you

Connect it to your operation

We would set it up with you: your ledger exports, your staging and reserving policies, your materiality, and the tie out run on every close. Reply wire it for a 30-minute slot.

Book a build call
data safety

Before you use live numbers

  • • Run last quarter's numbers first. Live data is not a test bed.
  • • Nothing here uploads to us. It runs in your own Claude account, on your own machine.
  • • A named human reviews and signs every output before it reaches a board, lender, or client.
  • • Wiring the open-source piece to real systems? Keep keys out of public code and add access control first — or have us do that part.
the fine print

Credit the original author

financeskills by Ayotunde Oso is MIT licensed (copyright KOMVIA) and used under its own license; the prompts are authored by consultance.ai. The skills draft journal entries for your controller to review and post; nothing posts itself. Nothing is hosted by us: your data goes only to your own Claude account.

Bring this workflow into your operation.

Book a build callBack to the library

Want this wired into your stack instead of running it yourself? That is our AI deal desk and finance automation service.

Newsletter

New builds, as they ship

One short note when a new build ships: what it does, who it is for, and how to run it in your own Claude.

Free. Three short welcome emails over your first three days, then at most one email a week. Unsubscribe in one click.

More like this

Related builds

All builds
Finance and data
Finance and data

Finance Skills for Claude Code

For CFOs, FP&A leads and family office analysts: install 7 open source finance skills in Claude Code for valuation, earnings prep and liquidity, with every input tied to the 10-K.

Setup guide
Finance and data
Finance and data

Month End Close AI Agent Team

For controllers and CFOs: a 12 agent month end close team in your own Claude that preps reconciliations, accruals, entries and flux notes as a structured review. You approve every ledger post.

Setup guide
Finance and data
Finance and data

Deal Red Flag Structured Review

For PE deal teams and family office direct investors: a structured review that sorts signed from draft, builds the evidence matrix, and gives every red flag one deal move, from reprice to escrow to walk.

Setup guide
Get the free kitBook a call

Share with your workflow owner.

Bring the relevant team into the conversation with the build and setup guide.

Forward by email
in one line

What is Quarter End Finance Skills for Claude?

Quarter End Finance Skills for Claude is a finance and data build in the consultance.ai AI Build Library. For bank, insurer and oil and gas controllers: 14 prompts that run 22 open source finance skills in your own Claude to draft quarter end entries, with a tie out gate. It fits Controllers and CFOs at banks, insurers and oil and gas producers who own the quarter end close: loan loss provisions, claims reserves, well retirement obligations and revenue. Setup difficulty is Medium, with 4 plain-English steps.

What does Quarter End Finance Skills for Claude do?

For bank, insurer and oil and gas controllers: 14 prompts that run 22 open source finance skills in your own Claude to draft quarter end entries, with a tie out gate.

Who is Quarter End Finance Skills for Claude for?

It fits Controllers and CFOs at banks, insurers and oil and gas producers who own the quarter end close: loan loss provisions, claims reserves, well retirement obligations and revenue.

How hard is Quarter End Finance Skills for Claude to set up?

Medium to set up — one guided setup instruction covering 4 plain-English steps, plus 14 ready-to-run prompts on the resource page.

How would consultance.ai build this out?

We would set it up with you: your ledger exports, your staging and reserving policies, your materiality, and the tie out run on every close. Reply wire it for a 30-minute slot.

What are the licensing terms?

financeskills by Ayotunde Oso is MIT licensed (copyright KOMVIA) and used under its own license; the prompts are authored by consultance.ai. The skills draft journal entries for your controller to review and post; nothing posts itself. Nothing is hosted by us: your data goes only to your own Claude account.

Connect this build to your workflow.

Discuss how Quarter End Finance Skills for Claude fits your systems, controls, and team. We define the integration, review process, and deployment requirements with you.

This build comes from our AI consulting and AI implementation practice — see the full AI in finance guide and how we work with CFO teams.

Book an audit