expense

A receipt and invoice reader that turns pasted text or a photo into one structured expense record, including the seller, date, currency, tax number, total, and item amounts.

In plain words
What is it for?
Use it to prepare expense claims or organise invoice data. It suggests a category but does not decide reimbursement or perform bookkeeping.
Why use it?
It reduces manual data entry and checks that the listed amounts actually add up to the stated total, even when scanned text contains common recognition mistakes.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/caprista/karvyloop/expense
Any agent
npx skills add Caprista/KarvyLoop --skill expense
Clone the repo
git clone --depth 1 https://github.com/Caprista/KarvyLoop

Made for: Claude Code, Codex.

Per session 165 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,086 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00165 $0.02086
Opus 5 $0.00082 $0.01043
Sonnet 5 $0.00033 $0.00417
Haiku 4.5 $0.00016 $0.00209

Measured 3d ago against content hash 71eff85dd9d5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

expense scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

karvyloop/system_skills/expense/SKILL.md · 126 lines

How it starts

The opening of the file, as written. The whole thing — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Expense receipt reading (system template)

A receipt is worth structuring only if the structure is trustworthy. The value here isn't "read some text off a picture" — it's turning a crumpled, mis-scanned slip into one line you can put in a claim without re-checking it by hand: right merchant, right date, right total, and — the part people skip — line items that actually add up to that total. This skill is a method, not a canned answer, and its scope is deliberately narrow: recognition and structuring, not judgment. It does not decide what your company will reimburse, does not submit anything, does not do your bookkeeping. It hands you a clean, checked record and a category hint — you stay the one who decides.

Honest input contract

This skill reads text or an image:

  • Text — paste the receipt's text, or the export from a bill/e-invoice. Cleanest path, always works.
  • Image — a photo or scan (jpg/png/…) is read via read_file, which OCRs it on this machine through the optional OCR extra (pip install "karvyloop[ocr]"; first use downloads a model, nothing is uploaded). If that extra isn't installed and the running model can't see images, the honest move is to say so and ask you to paste the text — never to invent a receipt from a filename. A vision-capable model may read the image directly instead; either way the same method below applies to the text.

The accountability rule

I (atom) answer to the role; the role answers to you. A structured line with a wrong total or an invented tax id is far worse than a field left blank and flagged. Money and tax numbers are unforgiving. When in doubt, leave it null and flag it — never a confident fabrication.

Procedure — do not skip steps

  1. Identify what the document is. receipt / invoice (发票, has a tax id and payee) / shopping list / itinerary / other. This decides which fields matter (an invoice's 税号 and 抬头; an itinerary's route and dates).

Read the full file on GitHub · 126 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 126 lines · 0 tokens per session scan A 71eff85dd9d5

Subscribe to this mod's changes

expense is a skill published in the GitHub repository Caprista/KarvyLoop (10 stars, last pushed 7d ago), licensed MIT. It adds 165 tokens to every session and 2,086 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.