# Originality and claim-boundary review ## Selected work - **Title:** *One Facility, Two Agents: The Command-Arbitration Layer for Self-Storage AI* - **Proposed slug:** `one-facility-two-agents-command-arbitration-self-storage-ai` - **Thesis:** Individually governed AI workflows can still issue stale, overlapping or incompatible commands against the same facility resource. A separate arbiter should require a typed command envelope, current-state precondition, explicit conflict policy, technical coordination, idempotent intent, bounded human handback and governing readback before consequential execution closes. - **Practical tool:** `agent-command-arbitration-checklist.csv`, a 20-control implementation and review checklist covering resource identity through closure. - **Fictional example:** Harbor Pine Storage, Gate 2, calendar versions 412 and 413, both proposed workflows, every actor, system, record, command, event, timestamp and outcome are explicitly fictional. ## Calendar, public corpus and active-work screen - The current governed editorial calendar at `/Users/jared/Documents/msu/authority-control-room/public/downloads/editorial-execution/editorial-calendar.csv` contained **39 data rows** and SHA-256 `3826428b30296902fcd5962fe14856def43aeae85e40ece4a78fae387d4b741d`. The older planning calendar contained 28 data rows and was also screened. - The current editorial-and-media register contained **60 data rows through MED-060** and SHA-256 `0ac0e4a4505a1d9d044e22de322eea0c6b137b56ea8edac1f73e90ff73b2bd9b`. The current submission tracker contained **287 data rows through SUB-287** and SHA-256 `1f9a3ff561480068ac3ac39a0f65f386fb5afae905c453ae8e7853b1a747ad8e`. - The live Jared authority sitemap returned HTTP 200 with **54 canonical locations** and SHA-256 `eb61346cedb20283f66851855b2fe401b9c4f65f7ea9e9061c3b05bbebacfbc0`. The live modSTORAGE post sitemap returned HTTP 200 with **128 post canonicals** and SHA-256 `7e2f9bb0eaf7dcc27d3acd66a5dc6c08f1d963c5f157a38fe0e8e61bf53f0c95` at the final pre-publication screen on 2026-08-30. - Today's morning lane covers first-hour water-intrusion response. Today's midday lane covers tested alternate ownership for critical facility functions. Neither title, thesis, slug, fictional example or practical tool overlaps this command-arbitration article. - Active, submitted, accepted, prepared and source-offer work screened included *The Event Bus Is Not the Source of Truth*, *From Dashboard Ranking to Defensible Facility Comparison*, *Retrieval Is Not Authority*, *The Oversight Load Test*, *The Facility Identity Cutover*, the CPE change-impact manuscript, the CREDA exception-carrying-cost manuscript, the BOMA Greater Los Angeles four-clock article and current AI-authority, observability, intervention, maintainability, continuity and transfer-risk source offers. - Exact local and live-sitemap searches found no selected-title, proposed-slug, Harbor Pine Storage or `agent-command-arbitration-checklist.csv` collision. - The closest owned article is *The Facility Intervention Contract*, which separates intent, authority, delivery, controller execution, physical effect and readback for one intervention. The new work addresses a different failure: two otherwise valid workflows competing for the same command domain before execution. - The event-spine paper governs transport versus observation, command and governing state. The operations-agent handback paper includes one compare-and-set race test. Neither supplies the new article's cross-agent collision classes, arbitration sequence, lease-versus-authority boundary, fencing epoch or 20-control operator checklist. - Current source offers about expiring action rights, physical-effect observability, maintainability and transfer risk were treated as allocated premises. This article does not reuse their central questions; it does not argue for expiring authority, prove physical effects, test technician maintainability or govern ownership transfer. ## Mechanical originality screen - Relevant corpus: **3,441 Markdown files** across the continual editorial execution and `/Users/jared/Documents/msu`, excluding the package itself, dependencies, Git data, build, distribution, render, archive, DOCX-QA and worktree paths. - Normalized body sentences containing 14 or more words: **84**. - Exact normalized 14-plus-word sentence matches: **0**. - Unique normalized six-word body shingles: **2,377**. - Maximum per-file six-word-shingle overlap: **0.126%** (**3** generic shingles), against the August 28 evening article *No Model Gets a Master Key*. - The three shared shingles are short architecture phrases about NIST AI RMF and conversion into a typed command. No sentence, paragraph, section, title, slug, thesis, scenario or tool was duplicated. This is a bounded local originality screen, not a commercial plagiarism-database certification or legal conclusion. ## Fictional content and claim boundaries - Harbor Pine Storage is entirely fictional. Every facility, gate, person, role, workflow, model, system, provider, record, version, policy, approval, command, lease, epoch, receipt, readback, timestamp, decision and result associated with the example is fictional. - The article and checklist do not report a real modSTORAGE, Facily.ai, customer, facility, employee, vendor, integration, product capability, deployment, pilot, performance result, service level, revenue, occupancy effect, adoption, partnership, certification, award, coverage or recognition. - AI output is treated as a proposal. It is never presented as observed facility fact, business authority, provider completion or physical outcome. - NIST sources are voluntary or broad government guidance, and AI RMF 1.0 is currently being revised. They do not prescribe or certify the proposed arbiter. - CloudEvents supplies event-identity semantics only. RFC 9110 supplies HTTP precondition and idempotency semantics only. Kubernetes supplies a lease implementation reference only. AWS documents a first-party API design approach only. OpenTelemetry supplies tracing semantics only. - Leases, entity tags, idempotency keys, traces and provider receipts are explicitly bounded from business authority, command compatibility and governing facility readback. - Actual access, life-safety, emergency, financial, customer, privacy, employment, contractual and legal decisions remain subject to qualified owners, approved policies, governing systems and applicable requirements. - The practical checklist records evidence references and states; it must not store passwords, recovery codes, customer data, restricted security details or secret credentials. ## Assistance, image and publication boundaries - The package was prepared with AI assistance for research synthesis, drafting and editorial QA. No legitimate authorship, source, image, ownership or assistance disclosure was removed or falsified. - Governed image 054 is disclosed as an AI-generated editorial illustration, not documentary evidence. It was copied byte-for-byte from the governed 3,840 × 2,560 master. - No author portrait was needed. The rejected black-turtleneck portrait was not used. - **Truthful state:** Publication-ready local draft only. No publisher handoff, submission, publication, canonical assignment, search submission, crawling, indexing, coverage or recognition occurred in this run.