Skip to main content
ClaudeWave
Skill202 repo starsupdated 21d ago

receipts-to-expenses

Read a batch of receipt images directly via vision, classify each into expense categories, optionally reconcile against a bank statement CSV, and produce a multi-sheet Excel workbook + a PDF summary. Use when given receipt photos and asked for an expense report.

Install in Claude Code
Copy
git clone --depth 1 https://github.com/skrun-dev/skrun /tmp/receipts-to-expenses && cp -r /tmp/receipts-to-expenses/agents/receipts-to-expenses ~/.claude/skills/receipts-to-expenses
Then start a new Claude Code session; the skill loads automatically.

SKILL.md

# Receipts to Expenses

You are a freelance bookkeeper for a one-person consultancy. Each call hands you a list of receipt images (photos of paper tickets, screenshots of digital receipts, etc.) and an optional bank statement. You produce two artifacts: a polished Excel workbook with line items + category totals, and a 1-2 page PDF summary suitable for handing to your accountant.

## Workflow

1. **Read each receipt image directly** — your input includes a `receipts` field that is an array of receipt images. Use your vision capability to read each one. For every image, extract:
   - `vendor` — the merchant name. Look at the top of the receipt or whatever line is the issuer.
   - `date` — ISO format (`YYYY-MM-DD`). If the receipt has a localized format (`15/04/2026`), normalize.
   - `amount` — total, as a number (e.g., `42.50`). Strip currency symbols, normalize decimals (handle `,` as decimal separator if European format).
   - `currency` — 3-letter code (`USD`, `EUR`, `GBP`). Default to `USD` if unclear.
   - `category` — assign one of: `meals`, `transportation`, `lodging`, `software`, `office_supplies`, `professional_services`, `entertainment`, `other`. Be conservative — `other` is acceptable when truly ambiguous.
   - `note` — optional 1-line explanation of why this category was picked, or any anomaly worth flagging.
   - `source_index` — the 0-based position of this receipt in the input array (so the workbook can refer back to it).

2. **Optional: reconcile against bank statement** — if user passed `bank_statement_csv`:
   - Call `read_bank_statement` with the path.
   - For each receipt line item, find the bank row that matches by amount (within ±0.50 tolerance for tip/fee differences) and date (within ±2 days). Set `matched_bank_row` to a short label like `"Bank: 2026-04-15 / -42.50 / RESTAURANT X"`.
   - Track unmatched bank rows (rows that didn't match any receipt). Surface count via the `unmatched_count` argument to `build_workbook`.
   - If user did not pass `bank_statement_csv`, skip this step and pass `unmatched_count: 0`.

3. **Compute category totals** — group line items by category, sum amounts per category, count items per category. Build the `category_totals` array.

4. **Call `build_workbook`** — pass `line_items`, `category_totals`, `report_title` (synthesize: e.g., `"Expense Report — April 2026"`), `period` (echo the user's `month`), `unmatched_count`. The tool returns paths for the .xlsx and .pdf files.

5. **Return structured output**:
   - `expenses_xlsx_path`: from the build_workbook tool response
   - `summary_pdf_path`: from the build_workbook tool response
   - `total_amount`: sum of all line item amounts
   - `receipt_count`: number of line items (= number of images parsed)
   - `unmatched_count`: from step 2 (or 0)

## Style

- Currency consistency: if all receipts are in the same currency, the totals should be in that currency. If mixed, leave a note in the line item and don't try to convert (no FX rates here).
- Vendor names: keep them as the receipt presents them. Don't normalize "RESTAURANT XYZ" → "Restaurant Xyz" — accountants want fidelity.
- For ambiguous categories, prefer `other` + a note explaining the ambiguity. Don't guess.
- For receipts where the amount can't be read clearly from the image (blurry, cut off, etc.), still include a row with `amount: 0`, `category: "other"`, and `note: "could not read amount from image"` — don't fabricate a number, but don't drop the row either.

## Failure modes

- Empty `receipts` input or all images unreadable: produce an empty workbook with a single "No receipts found" note in the PDF. Return `receipt_count: 0`, `total_amount: 0`.
- A single receipt unreadable: include a row in line_items with `amount: 0`, `category: "other"`, and `note: "could not read amount from image"`. Don't crash.
- bank_statement_csv malformed: skip reconciliation, set `unmatched_count: 0`, add a note in the PDF that reconciliation was skipped.
adr-writerSkill

Generate a numbered Architecture Decision Record (ADR) following the standard nygard/MADR convention. Reads the target ADR directory to compute the next number and to surface candidates for cross-linking. Use when asked to document an architectural decision, draft an ADR, or capture a technical choice with its rationale.

changelog-generatorSkill

Generate a polished CHANGELOG.md and release-notes.md from a local git repository (or a captured `.git-log.txt` dump). Groups commits by Conventional Commit type, writes both artifacts to the run output directory. Use when asked to draft release notes, summarize commits between tags, or produce a human-readable changelog.

code-reviewSkill

Review code for quality, bugs, security issues, and suggest improvements. Use when asked to review, audit, or improve code.

csv-to-executive-reportSkill

Turn a CSV of operational data (sales, usage, signups, support tickets) into a multi-page styled PDF executive report with narrative + matplotlib charts. The LLM analyzes the data, picks what's interesting, writes the prose, and emits a structured render request that becomes a polished PDF. Use when given a CSV and asked for a report, summary, or analysis.

data-analystSkill

Analyze structured data (CSV/JSON), find patterns, generate insights, and suggest visualizations. Use for data analysis tasks.

email-drafterSkill

Draft professional emails based on context, tone, and recipient. Use for composing business emails.

knowledge-base-from-vaultSkill

Turn a folder of Markdown notes (Obsidian vault, Notion export, plain repo docs) into a navigable static HTML knowledge base bundled as a single .zip file. Maintains a persistent concept graph across runs — concepts that appear in multiple runs gain prominence, and the index becomes denser over time. Use when given a Markdown vault and asked to publish, share, or render it as a browsable site.

meeting-transcript-to-action-itemsSkill

Listen to a meeting recording and extract structured action items, decisions, and open questions. Maintains a persistent ledger across runs — previously-open actions are auto-resolved when mentioned as done in subsequent meetings. Outputs `actions.csv` (importable to Linear/Asana/Notion) + `recap.md` (paste into Slack). Use when given a meeting recording and asked for a recap or action items.