How to process contractor invoices
Sheila uses written workflows to chain small accounting scripts, with its creator watching invoice processing, email drafts, and payment preparation in the terminal.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Process [contractor invoices] using [workflow instructions] and [available single-purpose scripts]. Draft related emails and prepare payments while exposing each action for [operator] review before anything is sent. Record bookkeeping updates through the supplied scripts.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- alexgleason
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The creator reports that Sheila handles the company's accounting flow, including invoices, payments, bookkeeping, expense tracking, and 1099s.
- Human involvement
- The creator observes invoice processing in the terminal and sees emails and payments before they go out, refining the workflow through repeated feedback.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- The creator replaced fragile multistep automation with more than fifty single-purpose scripts.
- They documented accounting workflows in a 600-line AGENTS.md file.
- Sheila read the instructions and chained scripts to process invoices.
Sheila is an open-source accounting agent that its creator says handles contractor invoices, fiat and Bitcoin payments, bookkeeping, expense tracking, and 1099s at their company. The creator first tried implementing complex multistep automation directly in code, but found that approach fragile.
The replacement architecture uses more than fifty granular scripts, each responsible for one operation. Examples include checking a balance, sending a payment, appending a spreadsheet row, and reading an email. A 600-line AGENTS.md file describes the workflows in plain English. Sheila reads those instructions and chains the scripts together to carry out accounting work.
The creator emphasizes close human feedback rather than unattended deployment. They sit in the terminal while Sheila processes invoices, watching as it drafts emails and prepares payments before those actions go out. They attribute reliability to hundreds of iterations between themselves and the AI. The post reports a working company accounting flow, but does not provide quantitative error rates or describe every approval mechanism. Sheila is published under the AGPL license.
Who it's for: Small-company operators comfortable supervising terminal workflows may find this script-based accounting approach relevant.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · alexgleason
Frequently asked questions
How does Sheila connect individual accounting operations?
The agent reads a 600-line AGENTS.md workflow document and chains more than fifty granular scripts, such as reading email, checking balances, and appending spreadsheet rows.
Does Sheila operate without supervision?
The creator describes sitting in the terminal while invoices are processed and seeing drafted emails and prepared payments before they go out.
More details
Remaining steps
- The creator watched drafted emails and prepared payments in the terminal before they went out.
About this account
Sheila chains individual scripts according to written accounting workflows while its creator watches execution.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47282367. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: alexgleason on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Sheila · AGENTS.md · Granular accounting scripts · Terminal
Accounting · Invoices · Human supervision
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.