Sheldon K.Salmon
I find the failure modes before they become legal exposure, reputational damage, or irreversible compliance gaps. Every document has a structural kill. The question is whether you find it — or your counterparty does.
Most document failure is not caused by careless authors. It is caused by authors reading their own work as supporters, not adversaries. The clause that seems clear to the drafter is exploitable by the party on the other side of the table. The obligation that seems enforceable in isolation becomes legally void when you trace its dependency chain to an undefined term in Annex II.
I read every document as the opposing party. Not to find rhetorical weaknesses — to find the structural failure modes: undefined thresholds, missing enforcement mechanisms, circular dependencies, obligations stated without triggering conditions. The findings come with ready-to-insert resolution language. Not a problem list. A corrected document.
I ran an independent adversarial structural analysis of the ISA Deep-Sea Mining Code — currently in active negotiation at the United Nations — on my own initiative, as a demonstration of what the methodology finds on a real high-stakes public document.
52 structural failure nodes across four sections. Four of them Critical — meaning the primary environmental protection obligation was legally unenforceable from day one. Not a legal opinion. A structural finding: the obligation existed in the text without the three conditions required for it to function as a legal obligation. The conditions were not in the document. The Annex that should have contained them referenced back to the main text, completing the circle.
Every finding came with ready-to-insert resolution text. Every resolution preserved the drafters' intent. The document could have been corrected before it entered its current negotiation phase. The findings have not been submitted to the Authority.
That analysis is the clearest demonstration of the methodology I can offer. It is publicly available. The document it analyses is publicly available. The findings are verifiable by anyone willing to read both.
Adversarial Reading First. Every document is read as the opposing party before it is read as anything else. The author's intent is secondary to the document's actual operative effect.
Structural Findings Only. Surface-level drafting errors, style inconsistencies, and rhetorical weaknesses are not the work. Structural failure nodes — the places where the document's own logic breaks — are the work.
Resolution Text Always. A finding without resolution language is half a deliverable. Every failure node ships with ready-to-insert language that closes the gap while preserving the drafter's intent.
Severity Must Be Named. Not all findings are equal. Critical gaps that make obligations legally unenforceable are not in the same register as minor drafting inconsistencies. The register difference must be explicit, not left to the client to infer.
Independence Is the Asset. An internal review team reads the document it helped write. An independent analyst reads it without the context that made the gaps invisible to the authors. Independence is not a credential — it is the mechanism of the method.
On the Instrument Stack
The methodology is not one instrument — it is a stack. The adversarial read identifies candidate failure nodes. The epistemic tagging protocol classifies each finding by confidence basis. The severity scoring system distinguishes Critical from Major from Minor by the downstream consequence of each gap activating. The resolution protocol closes each gap with language that survives the same adversarial read that found it. The stack is formal, versioned, and applied consistently across every engagement. This is why findings from a governance framework audit and findings from a technical specification audit can be presented in the same register and compared meaningfully.
Document Red-Team Passes. Adversarial structural analysis using the full instrument stack. Entry point: Targeted Pass (one section, one concern, one finding). Maximum: Comprehensive Adversarial Review (multiple independent passes, cross-instrument convergence testing, propagation analysis).
Framework Design and Audit. Purpose-built analytical frameworks for clients who need their own internal review capability. The same method I apply to documents, applied to the client's existing review framework — to find where the framework has coverage gaps, detection-claim failures, or self-validation that undermines its own findings.
Resolution Architecture. Every finding ships with repair language. This is not a separate service — it is built into every engagement. A findings register without resolution is a problem list. Problem lists generate anxiety. Resolution text generates corrected documents.