ai-ledger

Workings Template

A starter kit for AI-assisted accounting workflows using the Workings Layer Method.

Clone or copy this folder at the start of each period. Fill in the placeholders. Run inside workings/ so version control and AI scope stay aligned.

What’s Included

File / Folder	Purpose
`CLAUDE.md`	AI instructions – rules, permissions, data masking, tone
`.gitignore`	Excludes `data/raw/` and Python artifacts from git
`plan.md`	Workflow plan template – objective, steps, sources, outputs
`status_update.md`	Session log – milestone tracking
`data/raw/`	Landing zone for legacy source files (not committed)
`data/processed/`	Cleaned and intermediate data
`scripts/`	Processing scripts
`outputs/`	Generated reports and deliverables
`evidence/run-logs/`	Audit trail logs
`skills/`	Reusable SKILL.md files

How to Use Each Year

1. Clone or copy

If hosted as a git repo:

git clone <your-template-repo-url> workings
cd workings

If using it as a folder copy, paste it inside your year folder and initialize a new repo:

cd workings
git init
git add .
git commit -m "Initial setup from workings template"

2. Fill in the placeholders

Open CLAUDE.md and update the Year-Specific Customization section at the bottom:

Period (year)
Preparer name
Engagement or entity name
Legacy data paths (read-only source file locations)
Legacy output paths (where final outputs get saved back to)

Open plan.md and fill in:

Objective
Source file locations
Steps for this period
Output destinations

3. Pull in source data

Run copy_legacy_data.py (in scripts/) to copy files from legacy locations into data/raw/. Fill in the legacy paths before running – they are placeholders by design.

See examples/legacy-data-import for the full script and documentation.

4. Start Claude Code from inside `workings/`

cd workings
claude

Launching from inside workings/ sets the AI operating boundary to this folder. It can read ../ paths to reference legacy files above, but cannot write outside workings/.

5. Commit changes as you go

Commit after each meaningful step:

git add .
git commit -m "2026-04-26: Completed GL reconciliation, output in outputs/"

Year-Over-Year Audit Trail

Because every year starts from the same template, audit comparison is straightforward.

To compare this year’s CLAUDE.md against last year’s:

# From this year's workings folder
git diff ../2025/workings/CLAUDE.md CLAUDE.md

Or compare both years back to the template:

git diff <template-commit-hash> CLAUDE.md

This answers “what changed from last year?” without manually reviewing two folders.

Folder Structure

workings/
├── .git/
├── .gitignore
├── CLAUDE.md
├── plan.md
├── status_update.md
├── scripts/
│   └── copy_legacy_data.py
├── data/
│   ├── raw/              <- not committed
│   └── processed/
├── outputs/
├── evidence/
│   └── run-logs/
└── skills/
    └── [skill folders go here]

Cross-Year Analysis

When a workflow needs to look across multiple years, create a separate workings/ folder at the parent level (e.g., Budget/workings/). Give it its own CLAUDE.md that explicitly names the year folders it reads from:

Budget/
├── workings/             <- cross-year analysis, its own git repo
│   ├── CLAUDE.md         <- names ../2025/ and ../2026/ explicitly
│   ├── scripts/
│   └── outputs/
├── 2025/
│   └── workings/         <- cloned from this template
└── 2026/
    └── workings/         <- cloned from this template

Keep the cross-year workings/ as a separate git repo. It tracks the multi-year analysis independently of the year-level work.

Legacy Data Import Example*

This site is open source. Improve this page.