I find what kills a document before it costs you. Contract. Policy. Spec. Framework. Filing. If it's a document, I red-team it — and I show you exactly where it fails, with a paper trail back to the document's own text.
I take no sides. I read what the document actually says.
Complex documents have failure modes their authors didn't see. Rigorous authors miss things too — nobody reads their own document as an adversary. I do.
- What I Do
- Proof of Method
- One of the Instruments — FSVE
- Services
- Sheldon K. Salmon
- DOI / Citable Record
- How to Engage
- Links
I build formal instruments — scoring engines, falsification conditions, structured epistemic tagging — that locate undefined terms, missing enforcement mechanisms, exploitable loopholes, and structural failure nodes in documents before they go into production.
This isn't a proofread. It's an adversarial audit with a scored, falsifiable output and a citation back to the document's own language for every finding.
Where this applies: contracts, governance frameworks, AI/data policy, technical specifications, compliance filings — anything where a missed clause or an undefined term becomes a liability later.
Every claim I make carries an epistemic tag, so you always know exactly how certain a finding is:
| Tag | Meaning |
|---|---|
[D] |
Data — directly observed, measured, documented |
[R] |
Reasoned — logically derived from [D] evidence |
[S] |
Strategic — directional claim about future action |
[?] |
Unverified — open question, contested, unknown |
Nothing exits without a tag. Nothing is presented as more certain than the evidence supports.
Two independent structural reviews, run on live, high-stakes documents, on my own initiative — no client, no fee. Both are public and verifiable.
Adversarial structural review of ISBA/29/LTC/8, the ISA's draft Regional Environmental Management Plan guidance, currently in active UN negotiation. 52 structural failure nodes identified across four sections, including four critical gaps that leave the primary environmental protection obligation unenforceable from adoption. Every finding traced to the document's own provisions — no invented mechanics — delivered with ready-to-insert resolution text.
Published: DOI 10.5281/zenodo.21326719
Structural review of the governing Data Protection Addendum for a major healthcare AI scribe platform. Traced voice-pattern data through 45 CFR 164.514(b)(2)(i)(P) — which names voice prints as federally defined biometric identifiers — against the contract's delegation of biometric consent, notice, and deletion obligations to the provider organization, and business-associate liability under the HIPAA Security Rule since the 2013 Omnibus Rule.
The finding: a structural gap between what the contract offloads and what the provider's own intake consent language was ever built to cover.
Every claim links to its primary source; unverified claims pushed by a source were explicitly declined rather than promoted. Independently reviewed and validated by a healthcare compliance executive.
FSVE — Framework for Structured Validity Evaluation is one of the instruments in the stack — built for a specific, narrow purpose: scoring how much confidence a document or claim can actually bear, on defined dimensions, against a stated validity threshold. It's not the whole method; it's the piece that turns "this looks fine" into a number you can argue with.
- State: M-STRONG convergence · 75+ validated FCL entries · 0.813 expected validity baseline
- Documented, falsifiable, and open for inspection — not a consultant's opinion dressed up as a metric
- One document, up to 10 pages
- 1–2 page report on 2–3 real issues
- Delivered within 72 hours
- Capped at 5 per month; paid work takes priority
- NDAs welcome — I'll sign yours, and I'll red-team it too. A weak NDA protects no one.
The free pass is a single lens from a much deeper adversarial process — one filtered instrument out of a multi-stage build pipeline used on paid engagements. What you get for free is real; it's also the shallow end. That stage alone runs the artifact through a few narrow checks: a scan for what's present that shouldn't be and what's missing that should be, a severity gate that keeps minor issues from being reported with the same weight as structural ones, and an independence check that stops the review from sharing blind spots with whoever built the document. Paid work goes several stages deeper.
Role: Red-Team Frameworks Designer · Structural Integrity Analyst Location: Evans Mills, New York ORCID: 0009-0005-8057-5115
FSVE and the surrounding certainty-scoring infrastructure were built over twelve months of focused solo work, applying one standard throughout: every claim tagged, nothing presented as more certain than it is.
Affiliations: ARIA — Massively Scalable Neurotechnologies for Human Health Programme (Teaming Programme, Applicant, 2026–present) · Foresight Institute, Computation & Health Extension community (2026–present).
23 works registered and publicly archived through Zenodo. Full framework, brain-architecture, and world-building repos exist behind the scenes and are cited here for the record — the work above is the front door.
| Purpose | Cite This |
|---|---|
| Framework depth / certainty infrastructure (FSVE) | AION-BRAIN |
| Brain architecture / master manifest | AGI |
| Provenance and permanence systems | Sovereign Trace Protocol |
| Applied crisis/community platform | VERITAS |
| Organization as a whole | ORCID 0009-0005-8057-5115 |
If you need a document, contract, spec, filing, or governance framework red-teamed — that's the front door. First pass is free: one document, up to 10 pages, 72-hour turnaround, capped at 5 per month. Send the document.
If you are a skeptic — good. Test the method. Find the gaps. Challenge the tagging with specifics.
Contact
Sites
Publishing
Community
Support
AION System · Sheldon K. Salmon — Document & Framework Red-Teaming · Structural Integrity Analysis I find what kills a document before it costs you. I take no sides. I read what the document says.