Executive Summary — Strategic Theory of the Case
The Hub-and-Spoke Oversight Architecture: A Federal-State Financial Crimes Blueprint & Open-Source Agentic Engine for the 120th Congress.
Version 0.2.0 · Legal research cutoff: November 5, 2026 · Interface update: November 6, 2026 · Proposed congressional operating period: 2027–2028.
The institutional problem
Modern financial relationships span legal entities, intermediaries, property, securities, payment systems, digital assets, and government decisions. A single transaction may be relevant to several legislative questions without satisfying any criminal statute. Separate committees can receive overlapping records, repeat interviews, miss contradictory facts, and apply inconsistent handling rules. A shared professional service can improve analysis and preserve institutional memory while each committee retains its constitutional, statutory, and procedural responsibilities.
The Central Legal Hub proposed here supplies counsel, forensic accounting, technical analysis, evidence custody, verification, and lawful referral review. Standing committees remain the nodes that authorize inquiries and compulsory process within their jurisdiction. Senate participation requires separate Senate authority. State attorneys general and district attorneys independently assess jurisdiction, legal elements, admissibility, immunities, remedies, and whether to proceed. Citizen researchers contribute public-source research through a separately governed open repository. They do not become congressional staff, prosecutors, or authorized recipients of restricted records by joining this project.
The design is institutional rather than person-specific. A research task begins with a testable legislative or factual question and records evidence that supports and weakens competing explanations. Ownership, a foreign investment, a policy change, an anonymous wallet, a meeting, or a payment is not by itself proof of corruption. Public documents must distinguish established facts, allegations in source records, analytic inferences, unresolved questions, and legal proposals.
The operating model
- Define the legislative question and issuing body's jurisdiction. Record applicable law and its effective date; obtain authorization from the actual institution before exercising governmental powers.
- Begin with lawful public sources and voluntary information. Use scope-limited compulsory process only through authorized officials, after considering alternatives and burdens.
- Preserve originals and provenance. Attribute each factual proposition to an identifiable source and pinpoint, with contrary evidence and limitations retained.
- Use a shared catalog with legally separated evidence compartments. Restricted tax information, SAR information, sealed records, grand-jury material, classified information, and confidential committee records do not become generally shareable because they can be indexed together.
- Conduct staff-led, rights-respecting fact development and independent forensic review. Publish supported findings with sufficient context for readers to evaluate their limits.
- Refer only materials that counsel has determined may lawfully be shared. The receiving prosecutor independently obtains necessary evidence and makes decisions; a referral is neither a charging instruction nor an adjudication.
- Retain corrections and changes in law. Close the feedback loop by updating methods and skills when a verified error or a new authority changes the analysis.
Essential corrections to common assumptions
| Assumption | Operational correction |
| Staff depositions first require a new 2027 power | The 118th and 119th Congresses already authorized substantial counsel-led deposition practice. Appendix A proposes durable and clearer procedures. |
| A House rule can merge all tax and SAR records and send them to states | A House rule cannot override federal statutory secrecy. Access and redisclosure require the specific legal route applicable to each item. |
| Third-party records defeat every executive objection | Custody alone does not resolve privilege, constitutional burdens, official-act immunity, or the special presidential-records analysis. |
| Civil enforcement can be made immediate by an internal rule | Prospective House authorization can remove an internal approval step. It cannot create judicial jurisdiction, a cause of action, standing, or a guaranteed court timetable. |
| A New York clearing-bank transaction establishes every state offense | Territorial jurisdiction and each statute's elements require separate proof. Incidental routing is insufficient as a universal theory. |
| State enforcement is always pardon-proof | Presidential clemency generally addresses federal offenses; state proceedings still face jurisdiction, immunity, preemption and state-law limits. District of Columbia offenses require distinct analysis. |
| Three agreeing agents prove a finding | Signed audits establish accountable review events. Shared errors, compromised sources, and false premises remain possible. Human review is required. |
| A 30-minute heartbeat is a universal interoperability standard | Scheduling and configuration are runtime- and version-dependent. The autonomous-network roadmap requires explicit operator setup and capability testing. |
The legal chapters and Appendices A–D supply the supporting authorities and proposed instruments for these corrections. They are models for adoption and professional review, not existing institutional acts. They cannot be served or filed as completed process without matter-specific authorization, identifiers, scope, dates, and signatures.
The public and governmental systems
The public system contains the manual, verified public-source citations, downloadable draft instruments, portable skills, schemas, bounded research tasks, and reviewed public derivatives. It may receive public research submissions. It must not accept confidential government evidence into GitHub issues, commits, pull requests, public API payloads, or ordinary collaborative document comments. Public Git history is persistent: removing a file later does not reliably retract its prior disclosure.
The proposed governmental system is a separate, institutionally authorized environment. Its deployment, personnel access, legal entitlements, classification controls, procurement, retention, breach procedures, and records disposition must be approved by the responsible bodies. This repository implements neither a governmental evidence vault nor official referrals.
What this release provides
This release provides complete legal and operational chapters; model House procedural text; a legislative-purpose memorandum; phased third-party document-demand modules; a state-referral dossier standard; a hierarchical agent skill library; an evidentiary-packet JSON Schema; contribution and issue templates; structural validation and adversarial tests; a House-first interactive PWA and static portal builder; and Appendix G's autonomous-network design and API contract. A local Ed25519 audit helper demonstrates digest-bound three-auditor receipt verification without publication power.
It does not operate a live task API, authenticate external agents, deploy MCP or A2A services, enroll Moltbook communities, conduct private financial investigations, serve process, or run continuous research. The roadmap identifies the implementation and institutional gates for those proposed capabilities. No factual investigation or criminal accusation is included in the example packet.
Immediate decision standard
The first institutional decision is whether participating committees will adopt a limited, funded shared-service charter with clear decision rights and independent verification. The first public-project decision is whether maintainers can sustain source checking, corrections, security review, privacy review, and contribution moderation. A smaller accurately maintained repository is preferable to an unreviewable volume of allegations. Scale should follow demonstrated quality, independent staffing, lawful access, and operational controls.
The success measures are supported legislative findings, reduced duplication, preserved counterevidence, timely lawful dispositions, corrected errors, and transparent limitations. Counts of subpoenas, graph edges, wallet clusters, or agent submissions are activity measures and do not establish investigative merit.