[first2mars]

Self-employed estimates can deduct an excessive solo 401(k) contribution or stop when a valid, reviewed SSTB retirement deduction needs owner attribution. Missing prior-year AGI can also silently select the lower safe-harbor rate. This fixes those paths while retaining source-reconciliation and final-filing guards.

Closes #170.

Changes

• Validate legacy solo 401(k) employee, earnings-based employer and annual-additions limits using the proprietor's adjusted earnings and same-person other-plan deferrals. Reject excess contributions instead of silently clamping them. Ordinary base-limit contributions remain permitted at older ages; actual unsupported catch-ups and ambiguous multiple legacy plans fail explicitly.
• Add owner/business/amount/source attribution for a qualified retirement deduction in the retained single-SSTB review. Reduce AGI/QBI once, including replay of the actual retirement plan. Preserve the existing half-SE-only route and reject contradictory reviews.
• Keep a single reviewed SSTB with retirement and complete dividends on the simplified QBI path below the threshold. Within the TY2026 phase-in range, compute the active-QBI minimum from eligible SSTB income without restoring a fully excluded business.
• Require prior-year AGI only when 100% versus 110% can change the regular Form 2210 target; retain the explicit annual-payment override and public filing status for the MFS threshold.
• Decode file URLs in the CLI verification harness so paths containing spaces work.

The new regression tests use independently chosen synthetic amounts and invalid test identifiers. They are under forms/f1040/nodes/..., so the standard forms CI gate discovers them. The source/contract note links the IRS worksheets and statutes. No benchmark expectations, tolerance, failure baseline, production network policy, IRS schema or filing-source guard is relaxed.

Verification

Tested commit: 488675e88736a11b1690455f6eb9bbfbc49e8d90, based on current main 3add1ba73e84897a7c1f5d955c83866836e2d7fc.

• 190 focused tests across the QBI, retirement and underpayment files: 0 failures.
• Full strict typecheck and module-graph preflight: passed (3,398 modules).
• test:unit: 383 tests reported, 0 failures, one unavailable WSDL-schema prerequisite skipped; gate passed.
• test:forms: 8,349 tests reported, 35 existing baselined private-asset failures, 264 prerequisite skips; zero unbaselined failures and gate passed. The failure baseline is unchanged.
• check:structure, check:sandbox, check:cases, check:oracle, check:mef, check:ty2026, check:cli and check:irs: passed.
• Golden-value audit: 137 cases, zero changes. CLI gate: 18 complete sample returns.
• Implemented MeF rule gates: TY2025 374 assessed/373 clear; TY2026 374 assessed/231 clear; existing blocker baselines retained and zero gate problems.
• Native macOS compiled-binary release smoke: passed, including 30 complete CLI scenarios with seven exact summary amounts each, local MeF export and the TY2026 draft PDF.
• Changed-code lint and whitespace checks: passed.

The TY2025 legacy benchmark still has 22 passing and 87 failing cases. Compared with unchanged main, its counts and all 87 failed-case numerical differences/errors/unsupported keys are identical. The benchmark does not pass overall, and no goldens were edited to make it appear green. Upstream has no TY2026 benchmark case directory; explicit TY2026 regressions and the TY2026 sample gate cover these paths.

Local broad checks were run from an exact-commit clean checkout with canonical temporary paths, public IRS templates and the required PDF tools. This avoids existing macOS URL/temp-path and missing-tool failures without changing the tests.

Review boundaries

Positive legacy solo 401(k) inputs now require an explicit reconciled earnings basis. The employer worksheet uses the maximum 25% plan rate; a lower actual plan rate still needs review. Catch-up and multiple-plan arrangements outside the represented facts remain unsupported, rather than assumed valid.

The basic vehicle worksheet issue is already fixed upstream in #164, so mileage behavior is unchanged here. A separate local payment-planning command is not part of this contribution.

This is a draft calculation/source-validation contribution. Full private-evidence PDF/XSD/ATS coverage, source-document authenticity, IRS acceptance and TY2027 support are not claimed.