close-schedules

Skill: close-schedules

Purpose: Generate the supporting schedules a controller re-derives every close and staples behind the balance sheet as tie-out evidence — fixed-asset depreciation, prepaid amortization, deferred-revenue recognition. Producing them is a commodity (a fixed-asset sub-ledger or an ERP amortization module already does it); this engine earns its place the way the rest of the controller-autopilot does — it is self-checking and blocking, so a schedule can enter the governed close (close-approval-workflow) as evidence rather than as an unverified attachment.

Engine: ../../scripts/schedule_engine.py (stdlib only, Python 3.8+).

When to use

• You are wiring the close cycle and need the depreciation / prepaid / deferred-revenue schedules that support the fixed-asset, prepaid, and deferred-revenue balance-sheet lines.
• You want a rollforward whose arithmetic is checked, not just presented.

The three schedules

Subcommand	Rollforward identity	Ties into
depreciation	beginning NBV + additions − depreciation − disposals(NBV) = ending NBV	NonCurrentAssets (net of accumulated depreciation, a contra)
prepaid	opening + additions − amortization = ending	CurrentAssets (prepaid expenses)
deferred-revenue	opening + billings − recognized = ending	CurrentLiabilities (deferred/unearned revenue)

The two disciplines that make it honest

1. Every schedule must tie. Each subcommand is a rollforward whose identity is beginning + movements == ending, asset-by-asset and in total. Depreciation carries three parallel rollforwards — gross cost, accumulated depreciation, and NBV — with the cross-tie NBV = gross cost − accumulated depreciation checked at both ends. --strict exits non-zero (rc5) on a schedule whose portfolio rollforward does not tie. A schedule that does not tie is not evidence — it is a plug wearing a schedule's clothes.
2. Straight-line, disclosed. Straight-line depreciation (full-month convention), straight-line prepaid amortization, and ratable revenue recognition, each with a final-period catch-up so the balance lands exactly at its floor (salvage for an asset, zero for a fully-amortized prepaid or fully-recognized contract). The method is stamped on the artifact. Anything a real close needs beyond straight-line — accelerated/MACRS tax depreciation, usage-based recognition, ASC 606 variable consideration, partial-period conventions — is out of scope and must be modeled explicitly, never assumed here.

Presentation sign (reuses the statement-engine convention)

Amounts carry the sign of the statement section the schedule ties into, not an account's normal balance — the same rule as statement_engine._present(). Gross fixed assets and prepaids are asset-section (debit-positive) positive magnitudes; accumulated depreciation is a contra-asset (credit-natural) presented as a positive number that reduces NBV within the asset section; deferred revenue ties into a liability section (credit-positive). This is what keeps the contra correct.

Invocation

python3 scripts/schedule_engine.py depreciation \
  --assets examples/fixed-assets-2026-06.csv --period 2026-06 --strict --out fa-rollforward.json

python3 scripts/schedule_engine.py prepaid \
  --prepaids examples/prepaids-2026-06.csv --period 2026-06 --strict

python3 scripts/schedule_engine.py deferred-revenue \
  --contracts examples/deferred-revenue-2026-06.csv --period 2026-06 --strict

Input CSV shapes (SYNTHETIC examples in examples/):

• depreciation — asset_id,description,cost,salvage_value,useful_life_months,in_service_month,disposal_month,disposal_proceeds (disposal columns optional).
• prepaid — prepaid_id,description,total_amount,term_months,start_month.
• deferred-revenue — contract_id,description,billing_amount,term_months,start_month.

--period is YYYY-MM. The output also carries a full month-by-month full_schedule per prepaid/contract for the whole term (drains to exactly 0).

Correctness discipline (from the FORGE red-team)

• The golden fixture examples/expected-schedules-2026-06.json is hand-derived from the source CSVs by independent arithmetic, NOT frozen from an engine run — so a bug cannot ship inside its own golden. Worked cases: a mid-life machine (24/60 months), an addition placed in service in the period (depreciates its first month), and a mid-life disposal (removes cost + accumulated depreciation, books a $2,000 gain on proceeds).
• The acceptance suite ../../scripts/test_schedules.py asserts each rollforward ties, matches the golden, and that --strict blocks (rc5) a non-tying schedule.

Coordinating the tax calendar

The recognition/depreciation cadence here is the book view. Where book and tax diverge (MACRS vs. straight-line, ASC 606 vs. tax timing) that difference feeds the ASC 740 provision and the tax calendar — coordinated (not determined) via ../../knowledge/tax-close-calendar.md and ../../templates/tax-calendar.md. Those are controller coordination aids; a tax determination routes to a licensed CPA / tax advisor.

What this is not

Not an audit opinion, a GAAP determination, or a tax-basis calculation — outputs are decision-support (see ../../CLAUDE.md §3). A schedule derived from a summarized input file is not a substitute for a fixed-asset sub-ledger reconciled to the GL, nor for a tax depreciation calc; those land with the ELT / sub-ledger tier.