Open Net Worth
Open Net Worth is an open-source personal AI wealth assistant you run yourself: one ledger for your whole financial life, worked by an AI agent you control.
翻訳待ち
FBAR Proof Kit: local, source-linked support packets from foreign-bank statements.
翻訳待ち
Loading the full README. The saved excerpt is shown below.
FBAR Proof Kit by Skills To Pay The Bills
Turn machine-readable foreign-bank statements into a local, source-linked,
review-ready FBAR threshold support packet.
The kit chains three focused Agent Skills: statement intake preflight, retained
year-end Treasury FX proof, and an FBAR threshold check. Statement parsing and
packet generation stay on your machine. The included demo uses only invented
statements and a frozen public Treasury response, so you can inspect the whole
workflow without sharing data or making a network request.Read the full README ↗Turn machine-readable foreign-bank statements into a local, source-linked, review-ready FBAR threshold support packet.
Run the demo · Install the kit · See the sample packet
The kit chains three focused Agent Skills: statement intake preflight, retained year-end Treasury FX proof, and an FBAR threshold check. Statement parsing and packet generation stay on your machine. The included demo uses only invented statements and a frozen public Treasury response, so you can inspect the whole workflow without sharing data or making a network request.
This is support software, not tax or legal advice. It does not prepare or file FinCEN Form 114 and does not decide every filing obligation. See DISCLAIMER.md.
| Review artifact | What it establishes |
|---|---|
| A one-page JSON/CSV/PDF-backed threshold summary from a reviewed 365-day ledger, sealed by a portable postflight manifest. | |
| The exact year-end rate, source URL, date, reciprocal, retained response, and provenance. |
The happy path processes four fictional quarterly COP statements and retains a
reviewed packet. The refusal path omits Q2 and proves the workflow returns
insufficient-records / not-determinable and refuses ledger confirmation. It
does not manufacture a confident annual answer from incomplete evidence.
Readable-but-unsupported layouts are reported as parser coverage defects rather than missing data. Reviewed evidence stays visibly classified as formal extraction, diagnostic reconstruction, or user/preparer attestation. If an annual maximum has no known date, the result remains an explicit lower/upper interval instead of inventing a day.
The diagnostic reconstruction lane supports separate Debit/Credit columns and single explicitly signed Amount/Valor/Monto columns, including equal opening/closing zero-activity periods. Unsigned movements and unreconciled checkpoints fail closed.
npx skills add zztimur/skills-to-pay-the-bills \
--skill statement-intake-preflight \
--skill get-year-end-fx-rate \
--skill fbar-threshold-check
The skills CLI discovers all six public root
skills in this repository. By default, that third-party CLI sends anonymous
install telemetry used by the skills.sh leaderboard. This project receives no
statement data or first-party telemetry. Opt out before installing with:
DISABLE_TELEMETRY=1 npx skills add zztimur/skills-to-pay-the-bills \
--skill statement-intake-preflight \
--skill get-year-end-fx-rate \
--skill fbar-threshold-check
For a telemetry-free, inspect-before-install path, download the versioned skill
or fbar-proof-kit ZIP and SHA256SUMS from the
latest GitHub Release.
codex plugin marketplace add zztimur/skills-to-pay-the-bills
codex plugin add fbar-proof-kit@skills-to-pay-the-bills
/plugin marketplace add zztimur/skills-to-pay-the-bills
/plugin install fbar-proof-kit@skills-to-pay-the-bills
The native fbar-proof-kit bundle contains exactly the same three canonical
skills named above. The top-level skill folders remain the source of truth; a
deterministic sync check prevents the bundled copies from drifting.
FBAR Proof Kit runs through an Agent Skills-compatible host, not through a
model alone. The host must be able to load SKILL.md instructions, read local
files, run Python 3.11–3.13 and shell commands, preserve artifact paths, and
pause at human review gates.
| Agent host | Model guidance | Current status |
|---|---|---|
| Codex | Use the current recommended Codex default; prefer the strongest available reasoning model for ambiguous statement review. | Native plugin package available; model-level validation pending. |
| Claude Code | Use sonnet for the synthetic demo and routine runs; prefer opus for ambiguous statement review. | Native plugin package available; model-level validation pending. |
| Qwen Code | Use its current tool-capable default. | Agent Skills-compatible format; validation pending. |
| Kimi Code CLI | Use its current managed default. | Agent Skills-compatible format; validation pending. |
| Other hosts and models | Use only a tool-capable model in a host with Agent Skills, local-file, Python, and shell support. | Unverified. |
This is the beginning of the README. Read the rest at the original. Read the full README ↗