# Result — USD SOFR documentation commission

**Channel:** out (normal completed documentation result).  
**Documentation scope:** fulfilled at executor readiness level; separate assessment and actual institutional authorization remain pending.  
**Output directory:** `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali`  
**Executor:** Codex subagent `/root/sofr_executor_01`; exact underlying model identifier is not exposed in this invocation and is unknown.  
**First observed current time:** 2026-10-12 14:29:08 UTC (10:29:08 America/New_York). This is the first observed execution time, not an invented dispatch timestamp.  
**Record creation time:** 2026-10-12T14:45:43.766203+00:00 (UTC).  
**Observed interval from first clock reading:** 995.8 seconds; does not claim a separately observed dispatch duration.  
**Execution limit honored:** work completed before the caller's 20-minute boundary and absolute 2026-10-12 10:50 ET stop; no continuation scheduled.

## Selected agreement and input identities

The convenient kernel URL redirected to `https://pkg.prose.md/releases/0.4.6-rc.1/core/README.md`. Its bytes match the exact selected pin. The invocation explicitly selects release **0.4.6-rc.1**, adopting Quant documentation and its nested model-description, numerical-evidence, claim-evidence and institutional-facts definitions, plus the supplied base and house requirements. Definitions were read from their own public origins; local caching did not relocate their dependency context. No package installation or package resolver invocation occurred.

The retrieved program origin and final URL are both `https://openprose.ai/demos/sofr/v1/program.md`. Its SHA-256 is `ad458104fe99c7e1b0683558dd4cc8843d4756ba902dd7b7b9534a96cca859f9`. Source revision is `01ae3bdac6f61ecbabcd72e0116ebb2860e93157`, input as-of 2026-09-18, spot 2026-09-22; institutional dossier revision HER-scenario-1, authored 2026-10-12. Historical numerical receipts dated 2026-10-02 are inputs to this new documentation run.

| Retained definition/program | Requested public origin | Retrieved SHA-256 |
|---|---|---|
| `inputs/kernel-alias.md` | https://pkg.prose.md/kernel.md | `4622c21aa110d36abc9262496f922e578ecd9eb20e87241b28b2060d06df36cb` |
| `inputs/kernel.md` | https://pkg.prose.md/releases/0.4.6-rc.1/core/README.md | `4622c21aa110d36abc9262496f922e578ecd9eb20e87241b28b2060d06df36cb` |
| `inputs/documentation.md` | https://pkg.prose.md/releases/0.4.6-rc.1/quant/contracts/documentation.md | `80e8b260d5cfbd87abf63a5d94ee3e33387645775054eab599beeeccef9e2072` |
| `inputs/model-description.md` | https://pkg.prose.md/releases/0.4.6-rc.1/quant/contracts/model-description.md | `8a48a2f1ff8bea197026b3ba433ebb36cf9cfeb739cfd35b2eddfe992fe0c574` |
| `inputs/numerical-evidence.md` | https://pkg.prose.md/releases/0.4.6-rc.1/quant/contracts/numerical-evidence.md | `f4c14d6118d1a1846c5e1d60468a1af33c8b603e2d07949ed55dd0dd3035bd94` |
| `inputs/claim-evidence.md` | https://pkg.prose.md/releases/0.4.6-rc.1/quant/contracts/claim-evidence.md | `b023aeb88dc481b6cf31aff630a7adab28e52979827f5c6ecfda8861063d207c` |
| `inputs/institutional-facts.md` | https://pkg.prose.md/releases/0.4.6-rc.1/quant/contracts/institutional-facts.md | `cfb18dc33211a4b511e8af40172596a4f9df592a3d99d41c93da877fd0b10ed1` |
| `inputs/program.md` | https://openprose.ai/demos/sofr/v1/program.md | `ad458104fe99c7e1b0683558dd4cc8843d4756ba902dd7b7b9534a96cca859f9` |

All six selected definition hashes match the bundle's `selected-dependencies.json`. The public files with declared entries match `inputs/SHA256SUMS`. The complete **61-record** acquisition inventory is `input-manifest.json`: requested and final addresses, actual retrieval times, local paths, SHA-256 values and declared-hash comparisons. This includes acquired supporting files, some cached for reference even where not substantively used; the citation audit identifies the sources actually relied upon. Hash checks establish byte identity under the public bundle's declarations, not model truth or general agent conformance.

## Delivered artifacts

| Artifact | Actual absolute path | SHA-256 |
|---|---|---|
| `document.md` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/document.md` | `16d7918bc07263bef7f90b9ed2aa9f407ff01a3a28da4c3085c081f87ddca2bf` |
| `reviews.md` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/reviews.md` | `17c4ea54770d26f7fe4c04cc0ce4fb2b942d253fcbe10382b67ddf70c264cde8` |
| `checks.json` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/checks.json` | `afb21d0b01a07b2f56eb6b07f7375a4e95fc6c6699e506b47461666db79dbd5d` |
| `input-manifest.json` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/input-manifest.json` | `01a1ddb18fb620fd285715c280a44927aacb281730f39a35e503f486008945f1` |
| `inspect_evidence.py` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/inspect_evidence.py` | `938bb2b61b37893e4c764b3910ddfeecaf5ef1bdcafe82834ad60c5c781ce061` |
| `complete_document.py` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/complete_document.py` | `c82910f680b68a7b9638ffd264b48f37a6bbf18aa489b40d33942f996cd1c53b` |
| `finalize_records.py` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/finalize_records.py` | `a76471f71ce68cf289b0682f2a7ea079ae021768a54e011ab9950471b8fc8e78` |
| `figures/fig_discount_factors.png` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/figures/fig_discount_factors.png` | `25479261e6d8a12aa06936b8df47b7d4a6fa827d62ab948fbb4a144340709013` |
| `figures/fig_forwards.png` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/figures/fig_forwards.png` | `618e106048b281bba00661225174bd3d87696804c7fcf5b3a17432e29467c56b` |
| `figures/fig_bump_5y.png` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/figures/fig_bump_5y.png` | `4ddbc761a4b8345dc23af3ffe774e7fe2ded88fce079d89a20c31ab51e938c8a` |
| `figures/fig_delta_ladder.png` | `/private/var/folders/hz/g34mr1l52fn8d1y65j6gmq_80000gn/T/openprose-sofr-20261012-r01-ifbmpali/figures/fig_delta_ladder.png` | `338447c00027d61cac32da9ae61cae48a5a571e0f540ed2056f7bd74d260bebb` |

`document.md` contains all twelve exact numbered house sections, front matter, equations, complete calibration evidence, signed and absolute residual comparisons, end-date bins, limitations/controls, assumptions, simulated institutional record, references and both required disclaimers. Its prose count excluding table rows/reference-list bullets is 7,195; the whitespace count including table markup and references is 10,572. The commission's length target is advisory; retained evidence tables account for the difference.

`reviews.md` contains four separate records: requirement-by-requirement section plan, all 34 principal/smaller choices, a performed claim audit with support/unchecked/unavailable statuses, and readiness tied to the exact document hash. It assesses report requirements against reports and document requirements against the document.

The companion `artifact-manifest.json` records paths, sizes and SHA-256 for all retained files except itself, including `result.md`. Its separate location avoids an impossible self-referential hash inside `result.md`. Acquired evidence remains under `inputs/`; four figures are copied byte-for-byte under `figures/` and referenced relatively for offline reading.

## Checks actually performed

- Read the program, pinned kernel, all five Quant contracts, base/house requirements, brief, sources, source manifest, provenance and fictional dossier. Resolved declared nested adoptions; no incompatible release detected.
- Acquired named public evidence and code with `User-Agent: OpenProse/0.1` on the tracked retrievals. Verified six declared agreement hashes and bundle-declared hashes; checked every cached file against the acquisition manifest. Read original source paths only as provenance; never accessed the private repository.
- Inspected the calculation functions to establish pricing formulas, date/calendar conventions, monthly/business-day grids, exact bump-window endpoints, sign conventions, solver thresholds, quantile rules and component dependencies. Original model scripts were **not executed or imported**.
- Ran `python3 inspect_evidence.py` in this directory. This uses only standard-library arithmetic over frozen CSV/JSON: all 22 A/B calibration errors satisfy 0.01 bp; 361 monthly rows have the specified bounds and maximum date; full-precision JSON DF difference times USD100m gives USD227,458.19731463946; all 7,482 retained implementation differences match, with exact max/mean; held-out cohort counts and six distance-bin medians/IQRs were checked.
- Performed additional inline standard-library arithmetic using Python 3.14.6: reproduced historical signed summary/nearest-index quantile values for all ten cohort/method combinations; confirmed upper-middle held-out medians A0.45994244105472537 bp and B0.4046620120479877 bp; confirmed B269/A189/two ties and aggregate sign-test arithmetic. The different ordinary two-center medians were recorded rather than silently substituted.
- Compared each of five stage-receipt script hashes with the acquired script bytes; all matched. These checks do not make the historical stage receipts current runs.
- Visually inspected all four copied figures for axes, units and correspondence with the source plotting code. Checked exact twelve headings, local image targets, absence of placeholders, 14 institutional gap markers, verbatim house disclaimer and complete verbatim provider notice. Checked citations/relationships and all principal/smaller choices against source locators; details are in `reviews.md`.
- Wrote the document and completed its evidence tables with `python3 complete_document.py`. Wrote this result/companion manifest with `python3 finalize_records.py`. File creation and arithmetic inspection are the only local execution effects; no financial calculation is represented as reproduced.

Machine-readable outcomes and current timestamps are in `checks.json`. Current documentation-check Python identity is 3.14.6; it is separate from historical model Python3.12.14. Current user/system/model identities not otherwise exposed remain unknown.

## Observed effects and failures

Effects were public HTTP retrieval, creation of local input copies, creation of the requested Markdown artifacts and supporting records/scripts, and local figure copies, confined to the caller's fresh directory. Initial duplicate scratch `kernel.md` and `program.md` copies were removed after verified copies under `inputs/` existed. No accepted shared state, source data, code, previous run, developer repository, prior answer, remote publication, account, message recipient, institution or trading system was changed.

The first web-open attempts returned tool-level “not accessible” responses for the two launch URLs. Available HTTP clients then retrieved both successfully, followed redirects and recorded their identities. This was a recovered acquisition-tool limitation, not a remaining execution failure. No unresolved fetch, hash mismatch, write or arithmetic-check failure remains.

## Evidence limits, unfinished work and continuation

No required documentation artifact or supporting report is left unfinished. The following substantive evidence remains outside or unavailable within this commission, and disclosure does not supply it:

- Independent assessment of the exact document bytes, and actual institutional validation, authorization and sign-offs.
- A licensed benchmark, real internal feed/entitlement lineage, actual exposures, actual inventory/tier/roles, operating control and monitoring records, approvals/conditions, implementation tickets and the other 14 marked real-world evidence slots.
- Multi-date, stressed/negative-rate, full risk-system and separately sourced schedule/calendar studies. One date and shared adjustment/schedules limit current comparisons.
- Original raw archives/snapshots and omitted trade tables needed for full raw-data reproduction. This selective bundle permits numerical inspection, not a complete offline model rerun.
- Primary-paper and external-source assertions independently checked against their originals. Those contents were not retrieved; the document preserves them as developer attributions with explicit audit status, rather than treating titles as support. The code and frozen tests are available for implementation-specific statements.

The final readiness conclusion applies to documentation scope only. A later assessor should use the exact document hash in `reviews.md`, the input identities in `input-manifest.json`, and the findings there. Any edits create new document bytes and require a newly identified assessment target. No current or future institutional approval is inferred.
