finance-automation-portfolio
CLAUDE.md16 starsChanged 44 days agoMITView on GitHub
Fifty-two deterministic finance/tax engines + a self-healing loop layer, all on fictional data (seeded where the engine declares it). One rule everywhere: no material output rests on a single model's word.
- Installs packages
In this repository15 other files it gives its agents.
- apSkill
- atlasSkill
- brainSkill
- brand-graphicsSkill
- build-loopSkill
- closeSkill
- linkedin-postSkill
- loopsSkill
- reconSkill
- shipSkill
- site-updateSkill
- surplusSkill
- tax-1065Skill
- triangulateSkill
- validateSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 7
Every engine is a self-contained package: monthly-close-automation/close_engine, cash-reconciliation/recon_engine, …
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Sophon Finance Systems — portfolio repo
Fifty-two deterministic finance/tax engines + a self-healing loop layer, all on fictional data (seeded where the engine declares it). One rule everywhere: no material output rests on a single model's word.
Orient in 30 seconds
- Every engine is a self-contained package:
monthly-close-automation/close_engine,cash-reconciliation/recon_engine,tax-surplus-engine/surplus_engine,partnership-1065-automation/partnership_tax,audit-automation/validation_engine,ai-validation-framework/triangulate,knowledge-brain-engine/brain_engine,finance-atlas,cash-management/cash_engine,accounts-payable-automation/ap_engine,property-tax-automation/proptax_engine,surety-bond-automation/bond_engine,insurance-allocation-automation/insurance_engine,coi-compliance-automation/coi_engine,expense-report-automation/expense_engine,labor-charge-automation/labor_engine,lien-waiver-automation/lien_engine,equity-waterfall-automation/waterfall_engine,proforma-integrity-automation/proforma_engine,gross-receipts-tax-automation/grt_engine,filing-calendar-automation/filing_engine,energy-credit-automation/energy_engine. - Five engines carry a
loop.py(observe → detect → remediate → re-verify → gate); each has its own gate policy and its verdict doubles as the exit code. - Tests run per package (like CI):
cd <package-dir> && python -m pytest -q.SWEEP=1unlocks the ~1.72M-case grids. Deps:pip install -r requirements.txtonce. - The marketing site is
docs/index.html(GitHub Pages frommain— merges go live).
Use the skills — don't re-derive
.claude/skills/ carries the distilled operating knowledge from prior sessions. Invoke
instead of rediscovering:
- Engines:
/close/recon/surplus/tax-1065/validate/triangulate/brain/atlas/ap - The loop family + methodology:
/loops/build-loop - Brand + comms:
/brand-graphics/linkedin-post/site-update - Releasing work:
/ship
House rules
- Fictional data only (seeded where the engine declares it) — the atlas test suite lints a hashed confidentiality deny-list.
- Deterministic core: integer cents, seeded generators, byte-stable outputs. New code must not break byte-stability.
- Engines refuse rather than fudge; loops never invent a number; validation never writes.
- Human gate: autonomy is always framed as governed (escalate/quarantine what can't be proven).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

