Files
wanderbooks/docs/spec/rules.md
T
swaitsandClaude Opus 5.5 b22a034746 MCP: short connect text, a handbook, tool annotations, safe local imports
- The 13K-character instructions (clients cut them short) become a
  connect text of at most 1,400 characters whose first 512 stand alone,
  plus books.handbook and books://handbook/{section}: the old guidance by
  section, with no helper commands, home paths, pull requests or
  file-reading suggestions (tool_text_lint).
- Every tool has a title and read-only/destructive hints; no
  requiresUserInteraction anywhere (owner decision 2026-10-16).
- books.import_local_file reads only regular files inside the import
  roots (the downloads folder by default, or --import-root), never the
  data dir, books dir, vault or the key's folder, no symlink escapes or
  hard links, and the file opened must be the file checked.
- The store lock is a tokio mutex taken off the runtime threads.
- get_info refreshes settings first; a bare books://digest reads.

Review fixes (P0 review, 2026-10-16):
- Fail closed: a tokio mutex never poisons and spawn_blocking caught the
  panic, so after a panic mid-mutation the server kept serving a Store
  behind its journal. A panic (the task, or while holding the store)
  now stops the shared store; every later request on every session is
  refused with a plain message; a restart re-verifies the journal. The
  join error goes through redact_refusal and never carries the panic's
  text.
- books.set_w9 w9_file_path read any path (a link named w9.pdf to the
  vault key was stored and get_w9 wrote it out): it now goes through the
  import roots' lockdown, its type judged on the canonical path.
- Sealed even inside a root: the key file in force, where a linked one
  leads, both folders, and the default key folder whatever the settings
  say. Refusals check the roots first, so outside them every path gets
  one message (no probing what exists); inside, other books-dir files say
  to copy them out first.
- The open is O_NOFOLLOW | O_NONBLOCK (a pipe swapped in after the check
  can't hang a thread) and must match the checked file's device and
  inode. New dependency: libc (unix only), for those two open flags,
  which std doesn't name.
- Default import roots are the downloads folder only (the plan); files
  with full TINs (get_w9, vendor_1099 include_tins, the CPA package with
  TINs) are never written into an import root, and the server refuses to
  import any file it wrote with TINs (by SHA-256).
- set_tin and set_w9 overwrite vault records: destructive_hint = true;
  annotations_complete pins the key tools' hints. The read-only hint is
  checked in behaviour on every spec_mcp call (journal and data-dir files
  unchanged).
- Text: the connect text names the settings file as the one local file
  to edit, says every entry balances on real accounts and names its
  party, says invoice/bill/payment/payroll tools post unchecked, rewords
  the exemptions, and shows the company name as one plain line; the
  handbook takes section "all", lists its ids in the schema and says to
  read contract first; books.settings drops the config.json path and no
  error names the data dir or the key file. tool_text_lint covers the
  whole tool JSON and resources and bans git/commit/WL-0001; tool
  descriptions over 700 characters are pinned to today's eight (P5).
- digest: an empty since= is bare; the default date is pinned on every
  day of a month.

Round-2 review fixes (2026-10-16):
- TIN files (B-26): the in-memory, per-session set of TIN-file hashes
  became one registry beside the vault (vault/tin-files.sha256, hashes
  only), shared by every session and every later server. It holds the
  W-9 set_w9 reads (recorded as read), get_w9 copies, 1099 worksheets
  with TINs, CPA packages with TINs and each file in one that holds a
  TIN; a file is recorded before it is written, and a failure refuses.
  import_local_file, attach_receipt and file_document (base64 was never
  checked) refuse a recorded file, and a text file (by type, name, or
  UTF-8/UTF-16 decoding) holding a full TIN. Not caught: an altered
  binary copy; the docs say so (P2/P7 remove the path tools).
- The text check is narrow (owner/engineer decision, 2026-10-16: bank
  exports are filed for reconciliation, and has_full_tin's "nine digits
  however separated" refused their routing numbers, long references and
  dates run into amounts). Text is refused when it holds (a) a TIN the
  vault holds for a party, anywhere, as nine digits, 3-2-4 joined by
  hyphens, spaces or dots, or 2-7 joined by a hyphen or space, compared
  on the server against the decrypted vault and never returned or
  logged (a vault that can't be opened leaves (b)); or (b) a stand-alone
  printed SSN (123-45-6789) or EIN (12-3456789) shape: no digit, and no
  hyphen or dot followed by a digit, beside it. Full-width digits and
  Unicode hyphens and spaces read as ASCII. has_full_tin still guards
  refusals and results (redact_refusal). Known limit, for P7: a refusal
  under (a) tells the agent a text it chose holds a vault TIN.
  New dependency: zeroize in ledger-mcp (already in the workspace
  through ledger-vault, same major version), so the vault's TINs decoded
  for the scan are wiped from memory on drop as the vault's own are.
- Mutation testing of the final P0 diff (docs/testing.md): tests for
  each law a survivor broke (handbook titles, a dropped default root, a
  .jpg W-9, the waker, the walk's folder flags against a link and a
  pipe, the 40-link limit, save paths refused as such, the package
  marking only its TIN files, a registry that can't be read, text types,
  the short receipt rule's counts and a 12-character code). Simpler
  code where only a race could tell: inspect places an existing path by
  canonicalize itself (no second comparison), read_checked's read limit
  is its one size check, openat fails on -1 as specified. The 25 MiB
  upload test removes its 50 MB of books however it ends.
- The GUI's stale-rule prompt no longer says to open a pull request: it
  asks the agent to find the figure from the rule's official authority,
  confirm it with the human, pass it explicitly, and tell the human so it
  can be reported upstream (as the handbook says); a lib test pins it and
  bans files, git, GitHub and pull requests from every attention prompt.
- Import roots (local_files.rs INVARIANT 0): XDG_DOWNLOAD_DIR="$HOME/"
  made $HOME (or /) the root. No root may be /, the home folder or hold
  it, or hold the books, data dir, vault or a key: a named one stops the
  server at startup, a default one is dropped; the default falls back to
  ~/Downloads only if that is a real folder other than $HOME; with no
  root, an import says so.
- The read is a component walk from / with openat(O_NOFOLLOW |
  O_DIRECTORY) per folder and O_NOFOLLOW | O_NONBLOCK for the file, so a
  folder swapped for a link after the check fails; no step may land on a
  protected inode (one renamed into a root); dev/ino and nlink stay. No
  /proc. Portable to macOS. (libc's openat, two small unsafe blocks.)
- One refusal for missing, dangling and outside paths: resolved()
  follows every link, dangling ones too, so a link's target existing or
  not no longer changes the answer.
- Sealing: the key files (in force, default) and their targets always;
  their folders only when not $HOME, not a root, not holding a root; and
  ~/.config/wanderbooks whatever XDG_CONFIG_HOME says.
- Agent text (I-12): RuleNotKnown names the authority (IRS, with its
  address) and says to tell the human, never a repo path, PR or GitHub;
  the digest's stale_rules drop the repo catalog and path for it; vault,
  outside-books, STOPPED and local-file refusals name no git, path or
  flag. tool_text_lint bans docs/, .md, /home, data dir, --data-dir,
  --import-root, vault.key, .jsonl, sqlite, .gitignore, github, pull
  request, and also lints every error template (scanned from source),
  built error texts, and the server's refusal constants.
- The receipt rule told (ReceiptPolicy::rule) is the one enforced: the
  exact threshold in the base currency (no "$", no rounding), required
  and exempt accounts, reversals exempt; the connect text fits 1,400
  characters by dropping reasons, then codes, never the rule.
- The connect text follows the parties setting; "offer books.digest"
  everywhere. shown_name drops Unicode format characters (bidi, zero
  width).
- read_only_tools_change_nothing compares file contents, the database
  by every row, the vault and the settings file (raw database bytes did
  not change either; rows are the stable comparison).
- import_local_file takes the name and type from the canonical path;
  the digest's journal repair says when the tail was set aside, not the
  journal's file name; "version-history messages" back in the TIN rule;
  openWorldHint false on every tool, pinned in annotations_complete.
- Outside the MCP crate, for the same texts: ledger-core's DatedRule
  gains `authority` (the statute or publication with its address) and
  RuleNotKnown carries it; ledger-store's "sqlite error" reads "database
  error"; ledger-config's ReceiptPolicy::rule takes the base currency.
  New dev-dependency: rusqlite in ledger-mcp (already in the workspace,
  same version and features as ledger-store), to read the database's
  rows in read_only_tools_change_nothing.
- The filing scan is no TIN oracle (tin_files.rs INVARIANT 4): text
  holding more than two TIN-shaped numbers (any rendering rule (a)
  reads) that end in a known TIN's last four (the books' TIN and W-9
  facts, via known_tin_last4, before the vault is opened; then the
  vault's own) is refused with TIN_PROBE, the same words whether or not
  one is the vault's TIN; only one or two candidates are compared
  against the vault. Tests: unit
  text_probing_for_a_tin_is_refused_alike_whether_or_not_it_holds_it,
  one_or_two_candidates_are_compared_against_the_vault; over HTTP
  text_probing_for_a_tin_is_refused_alike_by_every_filing_tool.

Plan: plan/served.md P0, section 6 "Immediate hole (P0)", 4.1, 4.3, 4.9.

Review round 3. S1: a panic stops the server for every session, so no
input may reach one. Dates are four-digit years (new ledger_core::
calendar, 0001-01-01..9999-12-31): the MCP's parse_date is strict
YYYY-MM-DD, periods are bounded, last_day_of_month returns a refusal
(server and store); schedules, assets, recurring entries, periods,
documents, disposals and entries refuse dates outside the calendar, and
replay refuses such journal dates. Every amount a command takes is at
most a line's (draft lines by size, document lines, payouts, payroll,
statements, shares, tax basis, proceeds, recurring templates), on
replay too, so sums are exact; the 1099 vendor part and the K-1
property share use checked products (TooLargeToCompute); a recurring
run makes at most 1,000 drafts. tests/hostile_args.rs drives every tool
with schema-derived hostile arguments and every resource with generated
URIs, asserting the server never stops; the audit's panics are pinned.
S2: a `.` or `..` path name is refused as NO_SUCH_FILE up front. S3:
NUL-interleaved UTF-16 without a byte-order mark is scanned both ways.
S4: resource-read refusals are TIN-redacted. S5: set_w9 records its W-9
only once the call passes every check, just before the vault write.
S8: idempotentHint only on verify, run_depreciation, run_schedules,
run_recurring, set_party_roles (a repeat leaves journal and vault
alone; spec_vault checks every candidate twice); every tool's hints
pinned in one table. C1: books.settings lists receipt_required_accounts
and receipt_exempt_accounts uncut. C2: the default company name reads
"a new company". C3: the parties rule names no-party reasons. C4/C5:
the handbook names the settings file only as books.settings' books_toml,
and restores "check books.digest and offer it". C6: an import root
inside the books, data dir or vault is refused at startup. The lint
catches `just` however it ends and where the books are;
digest_resource_bare_reads reads the server's month.

Review round 4 (2026-10-17): round 3's bounds ran in replay too, so
journals main wrote and reopened (an entry dated 0000-06-01 or
+10000-01-01, a draft line of 2e15, an asset ending past 9999, a basis
or share past the line limit, sub-cent document and payroll lines) no
longer opened. Posted is forever: the bounds are the commands' only.
- Replay applies main's rule exactly: parse_date takes any date chrono
  reads; replay_add_line(_fx) has no size bound (add_line/add_line_fx
  do); replay_register_asset/correct_asset (chrono overflow only),
  replay_set_asset_tax_basis (check_journaled), replay_set_entry_parties
  (check_shares_journaled), replay_register_recurring
  (validate_recurring_journaled), replay_register_schedule (months 1 to
  600), replay_create_period (order only), replay_dispose_asset, and no
  line, date, payment or payroll range checks in replay_invoice/bill,
  their payments and replay_payroll_run.
- Derived figures over such books refuse (TooLargeToCompute,
  DateOutOfRange), never panic: amortization plans, depreciation into
  chrono's last month (and the true-up for lives past 9999), intervals,
  document totals and aging, net pay and payroll plans, money flows and
  share resolution, the asset register, receipt-policy debits; the MCP
  tools and the viewer show what can't be computed.
- tests/fixtures/old-books: the review's 13 journals, its control and 15
  more edges, written by main's Store API with main's figures
  (expected.json); every one opens in the writer and the viewer to those
  figures, every derived figure computes or refuses, every MCP tool and
  GUI page answers; hostile_args restarts the server after each
  property; each bound has a command/replay test pair.

Final review: a schedule's plan refuses a total of 10^26 or more
(TooLargeToCompute) instead of letting Decimal round its cents silently,
and a null `remaining` in the schedule output is always explained by
plan_error. Replay is unaffected: no journaled schedule month can belong
to a total that large (the line limit bounds every posted month).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01T3gXMVt7XJPi6ens1or9Dj
2026-10-17 09:25:00 -06:00

36 KiB
Raw Blame History

Rule catalog

Every rule the engine enforces, the authority for it, and what checks it. Status is AUTHORITY (a source states the rule), CONVENTION (common practice no authority states; the best reference is given), or DECISION (a human chose, and when). Vendored sources are in sources/; FASB and OpenStax are quoted briefly under their copyright and license (see README.md). Quotes from FASB's Codification were read in Deloitte's DART roadmaps, which reproduce it ("FASB Accounting Standards Codification: Copyright © 2005 - 2026 by Financial Accounting Foundation").

Spec blocks in the code cite these as SOURCE: rules.md B-n.


B-1 Every entry is a double entry: debits equal credits

Status: AUTHORITY

  • Pacioli ch. 36 (sources/pacioli-1494.md): "All entries made in the ledger have to be double entries — that is, if you make one creditor, you must make some one debtor."
  • IRS Pub. 583 (sources/irs-records.md): "It is self-balancing because you record every transaction as a debit entry in one account and as a credit entry in another. Under this system, the total debits must equal the total credits after you post the journal entries to the ledger accounts."

Enforced: kernel::check_lines via JournalEntry::validate (two or more lines, debits = credits exactly). A line's account must exist and take postings when the line is added, not only at post: DECISION (Stephen, 2026-10-13), Ledger::add_line (a_line_on_an_unknown_or_placeholder_account_is_refused_when_added). Checked: Kani check_lines_is_exact, accepted_entry_nets_to_zero; Lean Accepted.net_zero; property prop_posted_entries_balance.

B-2 A line's amount is positive; its side gives the direction

Status: CONVENTION

Pacioli keeps amounts in a debit column and a credit column (ch. 36: "All the creditors must appear in the Ledger at the right hand side, and all the debtors at the left"); the side, not a sign, says which way money moves. No authority forbids a negative amount outright; wanderbooks refuses them so that every line reads one way (2026-10-13, found by the formal pass).

Enforced: kernel::check_lines (NegativeAmountLine, ZeroAmountLine). Checked: Kani check_lines_is_exact; test negative_amount_line_is_rejected.

B-3 One money through the ledger; foreign amounts at the transaction-date rate

Status: AUTHORITY

  • Pacioli ch. 36: "the money that you reckon by at the beginning should be the same all through the Ledger."
  • ASC 830-20-30-1 (cited): "each asset, liability, revenue, expense, gain, or loss arising from the transaction shall be measured initially in the functional currency of the recording entity by use of the exchange rate in effect at that date."

A foreign line is booked in the base currency's smallest unit (ISO 4217 minor units, sources/iso-4217.md), half away from zero; a ledger in one money holds no fractions of its smallest unit, and its cash can then match a bank statement: DECISION (Stephen, 2026-10-13), after the GnuCash cross-check found wanderbooks kept $108.34567 where GnuCash booked $108.35. Lines booked before then replay as booked.

Enforced: Ledger::add_line (base currency only, no fraction of the minor unit), Ledger::add_line_fx (base amount = foreign amount × rate, rounded to the minor unit; original kept), Ledger::post_entry (every line, whatever made it: payroll, closes, disposals; FinerThanMinorUnit, since 2026-10-13 after a payroll run posted $65.27794). Journals from before replay as booked (replay_post_entry). Checked: a_foreign_line_is_booked_in_whole_base_units, no_entry_posts_a_fraction_of_the_minor_unit; the GnuCash oracle. Checked: foreign_currency_line_is_rejected, fx_overflow_is_an_error.

B-4 Five elements, closed list

Status: AUTHORITY (definitions); the five-type chart is CONVENTION

FASB Concepts Statement No. 8, Ch. 4 (2021), which replaced Concepts Statement 6 (cited): E16 "An asset is a present right of an entity to an economic benefit." E37 "A liability is a present obligation of an entity to transfer an economic benefit." E61 equity is "the residual interest in the assets of an entity that remains after deducting its liabilities." E80 revenues "are inflows or other enhancements of assets … or settlements of its liabilities … from delivering or producing goods, rendering services, or carrying out other activities." E81 expenses "are outflows or other using up of assets … or incurrences of its liabilities …".

Enforced: AccountType (five variants). A child account has its parent's type: DECISION (Stephen, 2026-10-13), Ledger::add_account (ParentTypeMismatch). Checked: a_child_account_has_its_parents_type.

B-5 Normal balances

Status: CONVENTION

No standard defines normal balances. OpenStax, Principles of Accounting, Vol. 1 §3.1 (cited; "Access for free at openstax.org"): "As assets and expenses increase on the debit side, their normal balance is a debit… Since liabilities, equity (such as common stock), and revenues increase with a credit, their 'normal' balance is a credit." Pacioli ch. 12: capital "must always be placed as creditor … and the cash always debtor."

Enforced: AccountType::normal_side, kernel::normal_balance. Checked: Kani normal_balance_is_signed, normal_side_increases; Lean balance_after_line.

B-6 The trial balance balances

Status: AUTHORITY

Pacioli ch. 36: "the total of the credits — I do not say creditors — should be equal to the total of the debits — I do not say debtors. If they were not equal there would be a mistake in the Ledger." IRS Pub. 583 (B-1).

Enforced: follows from B-1. Checked: Kani trial_balance_columns; Lean trial_balance_iff, Reachable.trial_balance, Reachable.trial_balance_through; properties.

B-7 The accounting equation

Status: AUTHORITY

CON 8 Ch. 4 ¶E62 (cited): equity "is the difference between the entity's assets and its liabilities." On a live book, income not yet closed is part of equity (B-11).

Checked: Kani accounting_equation; Lean accounting_equation, Reachable.accounting_equation[_through]; property prop_balance_sheet_balances.

B-8 Posted is forever: the book is written only forward

Status: AUTHORITY

  • Pacioli ch. 36: "You should not write backward, but always forward — that is, go forward as the days go, which never come back. If you do otherwise, the book would be untrue."
  • German Commercial Code (HGB) §239(3) (cited; official English translation): "An entry or a documented record may not be modified such that it is no longer possible to establish its original content."
  • Rev. Proc. 98-25 (sources/irs-records.md): machine-sensible records must keep the audit trail.

Enforced: posted entries immutable; the journal is append-only and hash-chained. Checked: Lean Step.extends, Steps.extends; property prop_posted_entries_immutable; store journal_tamper_is_detected.

B-9 A mistake is corrected by an opposite entry, never erased

Status: AUTHORITY

Pacioli ch. 31 (the storno): "you shall make another entry opposite this one in the credit for the same amount … After you have made this correction it is just as if you had written nothing in the debit column." Ch. 36: "Mark with a cross or with an "H" that special entry, and then make an entry on the opposite side under the same account."

Enforced: Ledger::reverse_entry (same lines, sides flipped, linked by reversal_of). Checked: Kani reversal_mirrors, reversal_cancels; Lean accepted_reversal_iff, movement_with_reversal.

B-10 An entry is reversed at most once; a closing entry never

Status: DECISION (Stephen, 2026-10-13)

Follows from B-9 (a correction corrects once) and B-11 (a closed period stays closed). Enforced: Ledger::check_reversal. Checked: an_entry_is_reversed_at_most_once, a_closing_entry_is_never_reversed.

B-11 Revenue and expenses close into profit and loss, and that into capital

Status: AUTHORITY (mechanics); the account the net lands in depends on the entity

  • Pacioli ch. 34: "you shall close this account into the capital account"; ch. 27: profit and loss "must then be transferred for its closing (saldo) into the capital account."
  • OpenStax §5.1 (cited): "Temporary (nominal) accounts are accounts that are closed at the end of each accounting period." Corporations close into retained earnings; partnerships into each partner's capital (B-13).

Enforced: Ledger::close_period_to. Checked: Kani closing_line_zeroes, closing_entry_balances; Lean closingEntry_zeroes, closingEntry_accepted, close_zeroes, close_credits_net_income; a_close_sweeps_only_through_the_period_end.

B-12 A closed period takes no new postings; periods close in order

Status: CONVENTION

Periods themselves: IRS Pub. 538 (sources/irs-periods-and-methods.md) defines the tax year, calendar and fiscal. Locking a closed period and closing in order are practice (and follow from B-8: a closed period's book is not rewritten). Periods nest (a month in a quarter in a year) and never partially overlap: DECISION (Stephen, 2026-10-13), Ledger::create_period (PeriodsOverlap; periods_nest_but_never_partially_overlap). Enforced: closed_period_containing, CloseOutOfOrder, Store::rehearse. Checked: Lean Steps.frozen; store property refused_commands_never_reach_the_journal (no draft in a closed period).

B-13 Partners share income by the partnership agreement; capital accounts carry it

Status: AUTHORITY (shares); DECISION (cent rounding)

  • 26 U.S.C. §704(a) (sources/irs-partnerships.md): "A partner's distributive share of income, gain, loss, deduction, or credit shall, except as otherwise provided in this chapter, be determined by the partnership agreement."
  • Form 1065 Schedule M-2: "Show what caused changes during the tax year in the partners' tax-basis capital accounts."
  • No authority says how to round a share to the cent. Largest remainder, ties to the earlier member: DECISION (Stephen, 2026-10-13). Closes made before then replay with the last member absorbing it.
  • Rounding never piles up: a run of members closes (same members and shares, back to back) splits cumulatively, each close giving a member their allocation of the run's total so far less their allocation before it, so each member's total from the run is within a cent of their exact share (follows from the running capital account below; found 2026-10-13: closing each month on its own gave every tied cent to the earlier member). Journaled as "rounding": "cumulative"; closes journaled "largest_remainder" replay one at a time.
  • A capital account is one running balance: Form 1065 Schedule K-1 instructions, Item L: "The partnership must report your beginning capital account and ending capital account for the year", "based on the partnership's books and records." So a statement of members' equity splits undistributed earnings cumulatively, as of each date (beginning as of the day before the period, ending as of its last day), never one period at a time: a member's ending on one statement is their beginning on the next, to the cent, and a member's period income is the roll-forward's remainder (their share within a rounding cent).
  • A correction undoes what it corrects (B-9): an entry and its reversal both in a statement's span are neither contributions nor distributions (reports::cancelled_in_span; the K-1 worksheet reads the same). The K-1 worksheet's contributed property (Item L) is what the member owned and handed over, with its own basis (§723): a prepaid service contract credited straight to capital is property (a contract handed over), while a prepaid service they paid a vendor for on the company's behalf, booked through Due to Shareholders (Dr prepaid / Cr 2100, then Dr 2100 / Cr capital), is cash (DECISION, Stephen, 2026-10-15); and it is shared among members credited in one entry in proportion to their credits (k1 INVARIANTS 3, 4, 6).

Enforced: CloseTo::Members, periods::allocate, kernel::apportion, members_equity. Checked: Kani apportion_is_exact, apportion_is_fair; Lean runPart_telescopes, closePart_total; properties prop_split_*, each_members_capital_chains_across_statements, each_member_stays_within_a_cent_of_their_share; a_run_of_closes_never_piles_up_rounding, members_closes_replay_under_the_rule_they_were_journaled_with; closes_journaled_under_the_old_rule_replay_with_it.

B-14 Statements: P&L for a period; cash flow by the indirect method

Status: AUTHORITY

ASC 230-10-45-28 (cited): an entity not using the direct method "shall determine and report the same amount for net cash flow from operating activities indirectly by adjusting net income … (the indirect or reconciliation method)." ASC 230-10-45-16: "All of the following are cash inflows from operating activities: Cash receipts from sales of goods or services…". A closing entry is not income or spending (B-11), so period statements exclude it.

Investing and financing are gross (ASC 230-10-45-7, read directly 2026-10-14, sources/fasb-asc.md: "information about the gross amounts of cash receipts and cash payments during a period is more relevant than information about the net amounts"; 45-8 allows netting only where "the turnover is quick, the amounts are large, and the maturities are short"). Financing inflows include "Proceeds from issuing equity instruments" and "other short- or long-term borrowing" (45-14(a), (b)); outflows include "distributions to owners" and "Repayments of amounts borrowed" (45-15(a), (b)). Every cash flow is investing, financing, or operating (45-10). Investing "include[s] making and collecting loans and acquiring and disposing of debt or equity instruments and property, plant, and equipment" (ASC 230-10-20). Which accounts hold loans made, investments, and debt is a fact of each company's chart, so it is a setting: [books] financing_accounts and investing_accounts in wanderbooks.toml (with cash_accounts), read by the MCP tools and the viewer alike.

Money an owner advances, by paying a company expense personally (a "Due to owners" liability), is borrowing: PwC Viewpoint, Financial statement presentation 6.9.3.1 (constructive receipts and disbursements) presents cash a third party pays on the entity's behalf "as though the transaction took place through the bank accounts of the reporting entity": a financing inflow and an operating outflow. Listing that liability in financing_accounts gives exactly that; repaying the owner is then a repayment of amounts borrowed. "Judgment is required" (PwC), so it is the company's setting, not a fixed rule.

Enforced: profit_and_loss_for_period, cash_flow. Checked: Lean statements_add_up; prop_period_pnl_partitions; cash_flow_reconciles_to_change_in_cash; financing_is_gross_borrowing_apart_from_repayment, a_loan_the_entity_makes_is_investing_gross, an_expense_the_owner_paid_is_borrowing_then_operating.

B-15 Depreciation: straight line, salvage never depreciated, partial months by days

Status: CONVENTION (book method); AUTHORITY for tax conventions, which wanderbooks does not apply

OpenStax §11.3 (cited): "Straight-line depreciation is a method of depreciation that evenly splits the depreciable amount across the useful life of the asset," with "(Cost – salvage value)/Useful life"; partial periods: "A common method is to allocate depreciation expense based on the number of months the asset is owned in a year." wanderbooks prorates partial months by days, including a schedule through a mid-month date (DECISION, Stephen 2026-10-13). ASC 360-10-35-4 (read 2026-10-14, sources/fasb-asc.md) prescribes no method: cost less salvage is spread over the useful life "in a systematic and rational manner," as equitably as possible to the periods served; straight line by days is one such allocation. Tax conventions (half-year, mid-month, mid-quarter): IRS Pub. 946 (sources/irs-depreciation.md), for the CPA.

Enforced: assets::depreciation_schedule. Checked: schedule_sums_exactly_to_depreciable_amount, salvage_value_is_never_depreciated, prop_depreciation_schedule_exact.

B-16 Prepaid expenses and deferred revenue are recognized over the period benefited or earned

Status: AUTHORITY (revenue); CONVENTION (prepaids)

ASC 606-10-45-2 (cited): payment before transfer → "the entity shall present the contract as a contract liability"; 606-10-25-23: "An entity shall recognize revenue when (or as) the entity satisfies a performance obligation by transferring a promised good or service." Prepaids: OpenStax §4.2 (cited); ASC 340-10-05-4 and 210-10-45-1(g), -2 (read 2026-10-14, sources/fasb-asc.md): a prepaid is an asset paid in advance of its use, current within the operating cycle; the ASC prescribes no expensing method, so monthly expensing stays a convention. Tax: the 12-month rule, Treas. Reg. §1.263(a)-4(f) (sources/irs-capitalization.md).

Enforced: schedules. Checked: a_prepaid_expense_is_expensed_month_by_month, deferred_revenue_is_earned_to_the_cent.

B-17 Platform payouts are recorded gross when the business is the principal

Status: AUTHORITY

ASC 606-10-55-37 (cited): "An entity is a principal if it controls the specified good or service before that good or service is transferred to a customer." 606-10-55-38: an agent "recognizes revenue in the amount of any fee or commission to which it expects to be entitled." 606-10-55-37B (read 2026-10-14, sources/fasb-asc.md): a principal recognizes revenue "in the gross amount of consideration." Whether an app developer is principal for app-store sales is a judgment under 55-39; the CPA confirms.

Enforced: payouts::payout_plan. Checked: a_payout_is_grossed_up_to_revenue_and_fees.

B-18 Aging buckets: Current, 1–30, 31–60, 61–90, 90+ days past due

Status: CONVENTION (bucket edges); DECISION (due today is Current)

ASC 326-20-55-37/38 (cited) illustrates "one way" to age trade receivables with buckets "current … 1–30 days past due … 31–60 … 61–90 … more than 90 days past due"; it is illustrative, not a rule. The FFIEC Call Report instructions (bank regulators, not binding here): "The past due status … should be determined in accordance with its contractual repayment terms." A document due today is not yet past due: DECISION (Stephen, 2026-10-13). Aging as of a date shows only documents issued by then, less only payments made by then; and a payment never predates its invoice or bill (a real prepayment is a deposit): DECISIONS (Stephen, 2026-10-13).

Enforced: arap::aging_bucket, Ledger::ar_aging/ap_aging, receive_payment/pay_bill (PaymentBeforeDocument). Checked: aging_boundaries, aging_is_exclusive_and_only_ages, aging_as_of_a_date_sees_only_what_existed_then, a_payment_never_predates_its_document.

B-19 1099 reporting thresholds

Status: AUTHORITY (sources/irs-1099-thresholds.md)

$600 through 2025; $2,000 for payments made after Dec. 31, 2025 (26 U.S.C. §6041(a), §6041A(a)(2), as amended by Pub. L. 119-21 §70433); for calendar years after 2026, indexed to inflation and rounded to the nearest $100 (§6041(h)). Card payments are reported on Form 1099-K, not 1099-NEC/MISC. Attorneys' gross proceeds stay at $600 (not modeled).

What counts as paid: payments made in the calendar year (§6041(a)), so a prepayment counts when made. The engine counts credits minus debits to the payment sources on a vendor's tagged entries: [books] cash_accounts and financing_accounts (a member or lender paying on the company's behalf; the company is still the payer). A card charge leaves neither (1099-K); a bill is not paid until it is (form1099::vendor_1099_report_with, checked by payments_come_from_cash_and_financing_only). A W-9 is on file by the vendor's flag or its party's W-9 facts.

An entry naming several parties (a payment shared between a contractor and a member; ledger_core::shares) counts only the vendor's share toward its 1099: what was paid to the payee (§6041(a)), not what the entry moved for others. The share is resolved (shares::resolve) against the entry's flow, the same target the MCP server checks shares against, then scaled to the 1099 payment: the share unchanged when the payment equals the target (the usual case, an expense paid from cash or by a member), else payment × share ÷ target, rounded to the nearest cent, half away from zero (so a part charged to a card scales out, 1099-K). A percent split's leftover cent goes to the party named first (shares INVARIANT 8). A reversal carries the shares and subtracts the same part, and re-apportioning either one re-apportions the other (Ledger::set_entry_parties INVARIANT 5), so the pair nets to zero. The vendor tag follows the parties: on an entry naming several, set_entry_vendor never replaces their shares (refused unless the vendor's party is named), and a vendor record never links to a party named beside another vendor (add_vendor), checked by tagging_the_vendor_keeps_a_shared_entrys_parties, clearing_the_vendor_on_a_shared_entry_is_refused, re_apportioning_a_reversed_entry_re_apportions_its_reversal, a_named_party_gaining_a_vendor_record_keeps_one_vendor_per_entry. Shares that don't resolve count the whole payment toward the vendor (the conservative answer) and the entry is listed (Vendor1099Report:: unresolved; a books.vendor_1099 note). An entry naming one party counts as before (form1099::vendor_1099_report_with_members, checked by a_shared_payment_counts_the_vendors_share, a_percent_split_gives_the_leftover_cent_to_the_first, a_reversed_shared_payment_nets_to_zero, a_shared_payment_partly_on_a_card_scales_to_the_payment, unresolved_shares_count_in_full_and_are_listed). DECISION (Stephen, 2026-10-15: shares).

Enforced: form1099::default_threshold, a dated rule (ledger_core::rules::THRESHOLD_1099, known through 2026, review from 2026-11-01). For a later year the engine refuses with RuleNotKnown rather than guess; the MCP digest and the viewer's "Needs attention" list flag the rule once it is due (DECISION, Stephen, 2026-10-13). Checked: an_unknown_year_threshold_is_never_guessed, dated_rules_go_stale_on_their_review_date, MCP stdio test.

B-20 Every transaction has supporting documents; records are kept

Status: AUTHORITY (sources/irs-records.md)

IRS Pub. 583: "In addition, you must keep supporting documents"; they "support the entries in your books and on your tax return"; keep records "until the period of limitations for that return runs out" (3 years generally; 7 for bad debts).

Enforced: the MCP receipt policy (not the engine; R-13).

B-21 Exact amounts in the books; rounding only on returns

Status: AUTHORITY (for where rounding happens)

Form 1065 instructions (sources/irs-partnerships.md): "If two or more amounts are added to figure the amount to enter on a line, include cents when adding the amounts and round off only the total." The books keep cents (and FX precision) exactly (R-2).

Enforced: Decimal money, Amount checked arithmetic, line range. Checked: Kani sums_are_exact; bridge tests.

B-22 Expense or capitalize (book policy, mirrors tax)

Status: AUTHORITY (sources/irs-capitalization.md)

Treas. Reg. §1.263(a)-1(f)(1)(ii)(D) sets $500 per invoice or item for a taxpayer without an applicable financial statement "or other amount as identified in published guidance"; Notice 2015-82 raised it to $2,500 for taxable years beginning on or after January 1, 2016. The 12-month rule: §1.263(a)-4(f)(1). The election is made yearly on the return.

Enforced: bookkeeper policy (books' AGENTS.md), not the engine.

B-23 Control roles: one account each, of the right kind, moved only when empty

Status: CONVENTION (a control account agrees with the records it controls); DECISIONS (Stephen, 2026-10-13 and 2026-10-15)

Invoices, bills, payroll, and closes post to accounts the engine picks by role: receivables, revenue, payables, general expense, retained earnings, salary and payroll tax expense, and five payroll withholdings. A role uses its default code (1200, 4000, 2000, 6999, 3999, 6100, 6110, 2110–2150) until the human assigns it one of the chart's own accounts; the assignment is journaled, so replay never depends on a setting outside the journal (DECISION, 2026-10-15). A role's account is postable and of the role's type (B-4), and the engine refuses one that isn't rather than renumber the chart (DECISION, 2026-10-13). No account fills two roles. A receivables, payables, withholding, or retained earnings account must keep agreeing with what it controls (open invoices, open bills, taxes withheld, closed earnings): such a role moves only when its account holds no balance, so the bookkeeper first moves the balance with an entry, which the journal shows. Revenue and expense roles move at any time; earlier postings stay where they are.

Enforced: Ledger::check_controls, Ledger::assign_role / check_assign_role, Ledger::check_renumber_account (an assigned role frees its code). Checked: ledger-core tests/spec_roles.rs; ledger-store replay of role.assigned.

B-24 Asset tags: one label per asset, never reused

Status: DECISION (Stephen, 2026-10-15)

Each fixed asset may carry a tag, the label stuck on the thing itself (generated PREFIX-0001 or the human's own: 1–32 ASCII letters, digits, - _ . /), and the manufacturer's serial number. Tags are unique ignoring case and are never reused by another asset, even after a retag or a disposal, so an old label can never point at the wrong thing. Both are journaled. The fixed asset register lists every asset as of a date and reconciles to the general ledger's cost and accumulated depreciation accounts.

Enforced: Ledger::set_asset_tag / check_asset_tag_for, assets::check_asset_tag, assets::check_serial. Checked: ledger-core tests/spec_asset_tags.rs; ledger-store tests/asset_tags.rs; ledger-mcp tests/spec_assets.rs.

B-25 Tax basis beside book value

Status: AUTHORITY (sources/irs-partnership-basis.md); DECISION (Stephen, 2026-10-15: the register carries tax basis)

Book value and tax basis are different numbers for the same asset. The book value of a member's contribution is its fair market value (Treas. Reg. §1.704-1(b)(2)(iv)(d)(1), cited). Its federal tax basis follows the rule that applies: property the company buys has a cost basis ("The basis of property shall be the cost of such property", IRC §1012(a)); property a member contributes keeps "the adjusted basis of such property to the contributing partner at the time of the contribution" (IRC §723, carryover); property a member converts from personal use depreciates from "the lesser of its fair market value or its adjusted depreciable basis" (Treas. Reg. §1.168(i)-4(b)(1)). The difference between fair market value and the contributor's adjusted basis is the built-in gain (positive) or loss (negative) that IRC §704(c) makes the partnership account for; a built-in loss counts only for the contributing member (§704(c)(1)(C)). wanderbooks records the rule, the contributor's adjusted basis, the resulting tax basis, and the built-in gain or loss; tax depreciation (MACRS, conventions, elections) stays with the CPA (sources/irs-depreciation.md).

Enforced: Ledger::set_asset_tax_basis / check_asset_tax_basis (cost takes no contributor basis, a contribution requires one, non-negative, in cents), Asset::tax_basis, Asset::built_in_gain, reports::fixed_asset_register (per-row tax fields, total_tax_basis, missing_tax_basis). Checked: ledger-core tests/spec_tax_basis.rs; ledger-store tests/asset_tags.rs (asset.tax_basis_set journaling, replay, v9 migration, corrupt payloads); ledger-mcp tests/spec_assets.rs.

B-26 Parties: one directory, no full TIN in the books

Status: AUTHORITY (truncated TINs: sources/irs-truncated-tin.md); DECISIONS (Stephen, 2026-10-15)

Everyone the company deals with is a party, with roles (vendor, contractor, customer, employee, member, payout platform, tax agency, financial institution, landlord). Vendor and customer records link to their party; entries are tagged with theirs. Names are unique among parties not archived, ignoring case and surrounding space. A W-9 is recorded as its facts and the TIN's last four digits (the part a truncated TIN shows, Treas. Reg. §301.6109-4(a)); no party field may hold a nine-digit number, and the full TIN lives only in the encrypted vault, outside the journal (DECISION). W-9 facts need a 1099 role (contractor, landlord); W-4 facts need the employee role; payroll lines that name their employee name a party with that role.

The full TIN, when the books need it for the 1099 worksheet, lives in the TIN vault (vault/ in the books dir, git-ignored; encrypted under a key file outside the books dir): books.set_w9 tin writes it there and journals only the last four and the vault record's hash; the signed W-9 goes there too, never to the attachment store. Any party may have its TIN in the vault, whatever its roles: every member's (for their K-1) and every 1099 payee's belongs there (DECISION, Stephen, 2026-10-15: "In the vault, that's the WHOLE reason we built it. And it needs to be in the damn CPA package."). books.set_tin stores one for any party; the books journal its TIN facts (type, last four, the record's hash: party.tin_set), which never disagree with its W-9 facts' last four. No tool output, error, or log holds a full TIN except books.vendor_1099 with include_tins (DECISION, Stephen, 2026-10-15: for the CPA, when asked): a refusal that quotes its input masks any TIN in it (parties::redact_full_tins, checked by refusals_never_echo_a_tin), and a party text field refuses nine digits however they are separated (space, hyphen, dot, slash, underscore; has_full_tin). Files holding one (books.get_w9, save_csv_to, and the CPA package, whose K-1 and 1099 worksheets carry every member's and payee's TIN by default) are written only outside the books dir and outside the server's import folders (DECISION, P0 review, 2026-10-16: so none can be filed back in as a receipt and read out again), and the package comes back only as its path. No file holding a full TIN is filed into the books (DECISION, P0 review round 2, 2026-10-16): the server keeps, beside the vault, the SHA-256 of every file it read or wrote with full TINs (the W-9 books.set_w9 reads, recorded as it is read; each books.get_w9 copy; a 1099 worksheet with TINs; a CPA package with TINs and each file in it that holds one), shared by every session and every later server for the books, and every way a file enters the books (books.import_local_file, books.attach_receipt, books.file_document) refuses those exact files, and refuses any text file (by its type or name, or because it decodes as UTF-8 or UTF-16) in which has_full_tin finds a TIN. What this can't catch: a changed copy of a binary file (a W-9 PDF re-saved or re-scanned), or a TIN in an image; plan/served.md P2 and P7 take the path tools and full TINs out of the agent's reach. The text scan reads nine digits with separators as a TIN, so a statement's text with a date run into an amount, or a long reference number, is refused too; such a file is filed as a PDF. A locked vault or a missing TIN never fails the package: it says which TINs are missing. books.purge_tin deletes a party's vault records and journals that it did.

Enforced: crate::parties (has_full_tin, check_text, W9Facts::check, W4Facts::check), Ledger::add_party, update_party, set_party_role, set_party_w9, set_party_w4, set_party_tin (TinFacts::check; TIN and W-9 facts agree), set_entry_party, validate_payroll; ledger-core k1 (the K-1 worksheet); ledger-vault (Tin::new, the encrypted records, records_held, purge_records); ledger-mcp books.set_w9, books.set_tin, books.get_w9, books.purge_tin, books.vendor_1099, books.export_cpa_package (open_vault refuses a key file in the books dir; outside_books for files with a TIN; vault_tin checks the vault against the books); ledger-mcp tin_files (the registry and the text scan, on every filing tool). Checked: ledger-core tests/spec_parties.rs; ledger-store tests/parties.rs; ledger-vault tests/vault.rs; ledger-mcp tests/spec_parties.rs, tests/spec_vault.rs (every response, the log, and every file in the books dir outside vault/ scanned for the TINs in any formatting), tests/tin_files.rs (a_tin_bearing_file_is_refused_from_every_session_and_after_a_restart), and the unit tests in tin_files.rs.

B-27 The journal format gate: newer is not corrupt

Status: DECISION (Stephen, 2026-10-16; plan/served.md 5.1 and P0); DECISION (P0 review, 2026-10-16: a line's own hash is checked before the line is called newer; the envelope and its hash are frozen across formats)

Every journal line has a type, and a version of wanderbooks reads a fixed set of them (ledger_store::format::EVENT_TYPES). A newer version may write a type an older one doesn't know, or declare a newer journal format with a journal.format_set {format, features} line before the first event that needs it. An older reader stops at the first line it can't read and says so: it never skips a line (the books it showed would differ from the record without saying so), and it never calls newer books corrupt.

What every journal format shares is frozen, so that an older reader can always follow the chain one line past the first line it can't read. A line is the envelope {seq, ts, type, payload, prev} and nothing beside it; its hash (what the next line's prev holds) is the SHA-256 of its canonical JSON: keys sorted at every depth, no whitespace. A newer format puts its new fields inside payload, never in the envelope, and never changes how a line is hashed. A line with a key beside the envelope's is refused as newer, never read with the key dropped, and its hash covers the key, so the next line still checks it.

Damage stays corruption. A line's seq and prev are checked before the gate, and its own hash is checked by the next line's prev; so when the gate refuses a line, the reader reads the next one first, taking only its prev (any JSON object with a string prev: it may use floats, a string seq or keys of a format this version doesn't read). If that prev doesn't match, the refused line was altered after it was written (a flipped bit in its type reads like an unknown type): corrupt. If it matches, the refused line is intact; the reader says so, and that it can't check the lines after it. If there is no next line, or it has no prev this version can read, nothing checks the refused line: the reader says it needs a newer wanderbooks or the line is damaged, and to run verify after updating, and never that it is intact. A newer format number is newer whatever the rest of its declaration looks like (a newer format may shape its features differently); a malformed declaration of a format this version reads is corrupt. A blank line, or one that isn't UTF-8 JSON, is damage, to the writer and the viewer alike: both read the journal as bytes split on newlines. A writer never declares a format it couldn't read back, or lowers the format in force. Opening the books checks every complete line before it changes anything: a cut-off last line is set aside only when the lines before it pass, so an older version never alters newer books.

Enforced: ledger_store::format (JournalFormat::admit, declared, gate, settle, declaration), journal::ENVELOPE_KEYS, JournalLine::parse, journal::read_journal_and_format, Snapshot::read_new, Store::open, Store::declare_journal_format, StoreError::NewerJournal (its verified flag picks the message). Checked: ledger-store tests/format.rs (old_reader_gets_newer_journal, a_damaged_line_that_looks_newer_is_corrupt_when_the_next_line_checks_it, an_unreadable_last_line_is_reported_as_newer_or_damaged, a_newer_line_the_next_line_chains_onto_is_newer_and_intact, a_newer_format_frames_its_lines_its_own_way, an_unknown_envelope_key_is_newer_and_checked_through, a_refused_line_followed_by_a_garbled_line_is_unverified, a_line_that_is_not_utf8_is_corrupt_for_the_viewer_and_the_writer, an_old_reader_never_repairs_newer_books, a_blank_line_is_corrupt_for_the_viewer_and_the_writer, damaged_journals_are_still_corrupt) and the unit tests in format.rs; ledger-gui tests/newer_journal.rs; ledger-mcp tests/startup.rs; the fuzz_store_replay target (the writer and the viewer agree on newer versus bad).