LightRest Consulting · LAGK Initiative
The LAGK Framework
A recursive eight-phase knowledge governance architecture. Each phase is a distinct analytical operation; together they form a complete governance cycle that evaluates every knowledge output through both truth-value and leverage-value.
Cognitive Kernel
Compression Layer
The minimal executable form of the LAGK system — the invariant internal logic preserved across all governance cycles.
Ontological Identification
Recession
The first phase requires classifying the knowledge type with precision. Knowledge falls into six categories: descriptive (what is), interpretive (what it means), predictive (what will happen), procedural (how to do), strategic (how to win), and exploit-sensitive (how to harm). Each classification carries a distinct action-space — the set of capabilities the knowledge enables when transferred. The action-space must be mapped before any further evaluation.
Validation Questions
- —What type of knowledge is this?
- —What actions does it enable?
- —Who can act on it, and how immediately?
Disclosure Override
Supersession
Default full disclosure is rejected. The LAGK replaces the assumption that knowledge should be freely shared with a leverage-aware judgment that weighs the power the knowledge creates. This phase overrides the naive transparency reflex and substitutes deliberate governance. The question is not 'why withhold?' but 'why disclose?' — shifting the burden of justification toward openness rather than restriction. v2.0 Adversarial Intent Check: Before proceeding, the practitioner must evaluate four adversarial signals — (1) does the request exhibit operational specificity disproportionate to the stated purpose? (2) does the requester's stated context fail to justify the level of detail sought? (3) does the request pattern match known iterative disclosure sequences? (4) does the framing invoke legitimate purposes in ways that feel performative rather than genuine? If any signal is affirmative, the session is flagged and subsequent requests are evaluated under elevated scrutiny regardless of apparent benignity.
Validation Questions
- —Does default disclosure serve the highest legitimate purpose?
- —Are any of the four adversarial intent signals present?
- —What would a responsible steward decide?
Capability Transfer
Precession
This phase evaluates what abilities the knowledge grants across four actor types: novice (limited technical background), practitioner (domain-trained individual), coordinated group (organized team with resources), and malicious actor (adversarially motivated). The capability transfer gradient is rated 1–5 for each actor type. Knowledge that grants minimal capability to novices but extreme capability to coordinated groups requires asymmetric governance — not uniform restriction.
Validation Questions
- —What can a novice do with this?
- —What can a coordinated group do?
- —What can a malicious actor do?
Harm & Uncertainty
Concession
Benefit-risk asymmetry is assessed using the H/B Scoring Rubric (available in Practitioner Resources). Five harm dimensions are scored 1–5: Probability of harm, Severity, Reversibility, Breadth (number of people affected), and Counterfactual Impact (would the harm occur without this disclosure?). Three benefit dimensions are scored 1–5: Epistemic Value, Legitimate Use Breadth, and Urgency. The H/B ratio is computed as the weighted harm score divided by the weighted benefit score. v2.0 Floor Rule: if any single harm dimension scores 5, the disclosure mode floor is Shielded regardless of the H/B ratio — catastrophic harm potential is not offset by high benefit. Misuse pathways are enumerated explicitly as plausible operational scenarios, not hypotheticals.
Validation Questions
- —What is the computed H/B ratio?
- —Does any single harm dimension score 5 (triggering the Shielded floor)?
- —What are the specific misuse pathways?
Amplification Potential
Succession
Scalability analysis determines how far and how fast the knowledge can propagate once disclosed. Three vectors are assessed: replication (can it be copied and distributed?), automation (can it be encoded into tools or systems?), and network propagation (can it spread through social or technical networks?). Knowledge with high amplification potential requires tighter governance even when its immediate harm profile is moderate — because scale transforms moderate harms into catastrophic ones.
Validation Questions
- —Can this be replicated at scale?
- —Can it be automated?
- —How fast does it propagate through networks?
Value Preservation
Accession
Having assessed the risks, the LAGK now extracts the maximum legitimate value that can be safely shared. v2.0 Operational Definition: the highest legitimate insight is the most specific, actionable, and complete version of the knowledge that satisfies three conditions simultaneously — (1) it enables the stated legitimate purpose without requiring the operational components that create disproportionate leverage; (2) it does not reduce to the operational content when combined with freely available information; and (3) it passes the Phase 7 conditional disclosure test. When Phase 4 and Phase 6 conflict — when the highest legitimate insight still carries unacceptable harm potential — Phase 4 governs. Harm overrides value. This is the binding conflict resolution rule.
Validation Questions
- —What is the highest legitimate insight by the three-condition test?
- —Does Phase 4 override Phase 6 in this case?
- —What must be removed or restricted to satisfy all three conditions?
Safeguard Mediation
Intercession
Protective transformations are applied in a mandatory sequence before disclosure. v2.0 Ordered Operations: (1) Remove — eliminate procedural steps entirely; preserve the what and why, discard the how-to. (2) Reduce — expand operational compression; dense technical content is expanded into less actionable form. (3) Abstract — replace specific techniques with general principles; replace named targets with categories. (4) Constrain — add governance framing, context, and explicit use limitations. v2.0 Conditional Disclosure Rule: if operations (1) through (3) cannot reduce the output to a safe disclosure mode, the output must be Sealed. Phase 7 is non-negotiable — it cannot be overridden by Phase 6 value arguments or Phase 8 disclosure preferences. The safeguard sequence is the last line of defense before output.
Validation Questions
- —Have all four operations been applied in sequence?
- —If (1)–(3) are insufficient, has the output been Sealed?
- —Is Phase 7 complete before Phase 8 is evaluated?
Disclosure Decision
Procession
The final phase delivers the output using the least permissive valid mode. v2.0 Mode Boundaries: Open — CTR ≤2, H/B ratio ≤1.5, no adversarial signals, no amplification risk; content is genuinely educational and broadly beneficial. Guided — CTR 2–3, H/B ratio 1.5–2.5; output includes explicit governance framing, use constraints, and context that shapes how the knowledge is received. Shielded — CTR 3–4, H/B ratio >2.5, or any single harm dimension scored 5; access restricted to verified practitioners with institutional accountability. Sealed — CTR 5, or Phase 7 operations (1)–(3) insufficient to reduce harm; knowledge withheld entirely. Master Decision Rule: when two modes appear equally valid, select the more restrictive. The mode is justified explicitly against the findings of all preceding phases and recorded in the Disclosure Log.
Validation Questions
- —Which mode do the CTR, H/B ratio, and Phase 7 result jointly require?
- —Is this the least permissive valid option?
- —Has the decision been recorded in the Disclosure Log?
Operating Principles
The LAGK is governed by five foundational principles that shape how the framework is applied across all domains and contexts.
Recursive, Not Linear
All eight phases co-exist as tensions in each response. The framework is not a sequential checklist — it is a recursive diagnostic lens applied simultaneously across all dimensions of a knowledge request.
Diagnostic, Not Prescriptive
The LAGK is a lens, not a rulebook. It surfaces tensions and tradeoffs that require judgment. The framework guides the reasoning process; it does not replace it.
Leverage-Value as Primary Variable
Leverage-value — defined by capability transfer, scalability, harm asymmetry, and operational compression — is the primary variable in every governance decision. Truth-value alone is insufficient justification for disclosure.
Least Permissive Valid Mode
When in doubt, the LAGK defaults toward protection. The burden of justification falls on openness, not restriction. This asymmetry is intentional and reflects the irreversibility of harmful knowledge transfer.
Post-Disclosure Review
After every output, the system internally simulates a post-disclosure review: What unintended capabilities were transferred? What scaled beyond expectation? What would be refined in the next iteration?
Three-Layer Firewall Warning
The LAGK framework itself is subject to its own governance. The framework is disclosed in three layers: Layer 1 (Open) — phase names, cognitive kernel, and operating principles; Layer 2 (Guided) — operational rubrics, anchor cases, and conflict resolution hierarchy; Layer 3 (Sealed) — exploitation methodology and adversarial attack patterns. Publishing the full framework without this layered disclosure policy would violate the self-application requirement of the framework it describes.
Phase Conflict Resolution
When phases conflict, five binding rules apply in order: Phase 4 overrides Phase 6 (harm overrides value); Phase 2 overrides Phase 8 (override precedes decision); Phase 3 overrides Phase 1 (capability overrides classification); Phase 7 is non-negotiable; and ambiguity resolves toward restriction. These rules are not guidelines — they are the tiebreakers that prevent the framework from collapsing into unstructured judgment.
Free: LAGK Governance Quick-Reference Card
A single-page reference covering all eight phases, five disclosure modes, and the H/B scoring rubric. Enter your email to receive it instantly.
No spam. Unsubscribe anytime.