Skip to content
Nexload SDK

Cognitive Graph Reasoning (Tier 1) | Nexload SDK

The Nexload Cognitive Graph separates six cognitively distinct jobs, orchestrated by a lightweight kernel. Rather than asking a model to “think hard” in an unstructured stream, each specialist is standalone-capable, possesses bounded input/output contracts, and adheres to strict operational boundaries.

The 6+1 Cognitive Graph Topology

Section titled “The 6+1 Cognitive Graph Topology”
               ┌─────────────────────────────────┐
               │        nexload-reasoning        │
               │   (Kernel, Router & Budget)     │
               └────────────────┬────────────────┘
                                │
        ┌───────────────────────┼───────────────────────┐
        ▼                       ▼                       ▼
 ┌──────────────┐       ┌───────────────┐       ┌──────────────┐
 │  Discovery   │       │ Investigation │       │   Ideation   │
 │ (Scope/JTBD) │       │(Empirical Bug)│       │ (Divergence) │
 └──────┬───────┘       └───────┬───────┘       └──────┬───────┘
        │                       │                      │
        └───────────────────────┼──────────────────────┘
                                ▼
                        ┌───────────────┐
                        │    Design     │
                        │(Arch/SSOT/Seam│
                        └───────┬───────┘
                                ▼
                        ┌───────────────┐
                        │  Evaluation   │
                        │(Pareto Pruning│
                        └───────┬───────┘
                                ▼
                        ┌───────────────┐
                        │   Execution   │
                        │(Atomic Verify)│
                        └───────────────┘

1. The Kernel: Router & Depth Controller

Section titled “1. The Kernel: Router & Depth Controller”

nexload-reasoning

Section titled “nexload-reasoning”

The cognitive kernel determines what reasoning is needed, at what depth, routes to specialists, and enforces stopping discipline.

  • Core Mission: Assesses task complexity (Low, Medium, High), provisions a cognitive budget, and halts reasoning immediately once remaining uncertainty has zero material impact on architecture or risk.
  • When to Use (Positive Trigger): Ambiguous requests, multi-step engineering initiatives, high-consequence refactors, or recovery from implementation deadlocks.
  • When NOT to Use (Boundary): Deterministic single-step edits, trivial formatting, syntax fixes, or when the request already cleanly maps to a single specialist.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning

2. The 6 Reasoning Specialists

Section titled “2. The 6 Reasoning Specialists”

Discovery Specialist (nexload-reasoning-discovery)

Section titled “Discovery Specialist (nexload-reasoning-discovery)”
  • Mission: Problem space framing. Extracts root user intent (Jobs-to-be-Done), separates requirements from solution hypotheses, locks scope (IN, OUT, MUST-PRESERVE), and gates user questions.
  • Positive Trigger: Feature requests disguised as implementation solutions (e.g. “Add Kafka” or “Migrate to Microservices”); greenfield initiatives.
  • Negative Boundary: Debugging known errors with existing stack traces; comparing pre-existing designs; routine coding.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-discovery

Investigation Specialist (nexload-reasoning-investigation)

Section titled “Investigation Specialist (nexload-reasoning-investigation)”
  • Mission: Reality and causal engine. Audits empirical evidence, forms competing hypotheses ($H_1, H_2, \dots, H_n$), designs decisive discriminator tests, and isolates verified root causes.
  • Positive Trigger: System crashes, intermittent panics, memory/CPU leaks, performance degradation, contradictory log statements.
  • Negative Boundary: Brainstorming new user flows; designing clean-slate schemas; planning sprint delivery.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-investigation

Ideation Specialist (nexload-reasoning-ideation)

Section titled “Ideation Specialist (nexload-reasoning-ideation)”
  • Mission: Divergence mechanism generator. Expands solution space across control, state, and subtraction axes; breaks mental fixation via assumption inversion under the Divergence Firewall.
  • Positive Trigger: Architectural fixation; exploring novel mechanisms; generating at least 3 distinct conceptual alternatives.
  • Negative Boundary: Choosing between Option A and Option B; reviewing pull requests; production incident recovery.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-ideation

Design Specialist (nexload-reasoning-design)

Section titled “Design Specialist (nexload-reasoning-design)”
  • Mission: Architectural synthesis. Synthesizes coherent system models, establishes Single Source of Truth (SSOT), deep module seams, failure path semantics (retries, timeouts, circuit breakers), and enforces complexity limit to Now + 1.
  • Positive Trigger: Structuring chosen concepts into deep modules; designing API/DB schemas; defining failure recovery behavior.
  • Negative Boundary: Unconstrained brainstorming; final commercial/vendor selection; writing routine implementation code.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-design

Evaluation Specialist (nexload-reasoning-evaluation)

Section titled “Evaluation Specialist (nexload-reasoning-evaluation)”
  • Mission: Convergent decision engine. Performs strict Pareto pruning of dominated options, weighs material trade-offs without scoring theater, runs pre-mortem stress tests, and recommends a single actionable default.
  • Positive Trigger: Choosing between competing designs or libraries; high-stakes migrations; vendor selection; pre-mortem audits.
  • Negative Boundary: Generating more alternatives; coding solutions; collecting initial stack trace evidence.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-evaluation

Execution Specialist (nexload-reasoning-execution)

Section titled “Execution Specialist (nexload-reasoning-execution)”
  • Mission: Delivery and claim verification. Decomposes atomic dependency-ordered plans, exercises controlled autonomy on routine details, proves completion with empirical evidence, and routes failures.
  • Positive Trigger: Implementing approved designs; executing migrations; structured refactoring plans; verifying deliverables.
  • Negative Boundary: Re-debating settled architecture; brainstorming features; guessing root causes of crashes.
Terminal window
npx skills add gecut/nexload-sdk --skill nexload-reasoning-execution

Capability & Trigger Matrix

Section titled “Capability & Trigger Matrix”
SpecialistCLI IdentifierPrimary ObjectiveKey Output Contract
Kernelnexload-reasoningRoute & Depth ControlCognitive routing decision & stopping check
Discoverynexload-reasoning-discoveryFrame Problem SpaceBounded scope (IN/OUT/PRESERVE) & JTBD
Investigationnexload-reasoning-investigationIsolate Causal RootDiscriminated hypothesis & empirical proof
Ideationnexload-reasoning-ideationDivergent Mechanisms3+ distinct approaches behind firewall
Designnexload-reasoning-designArchitectural SynthesisSSOT data flow, deep module seams, failure modes
Evaluationnexload-reasoning-evaluationConvergent DecisionPareto frontier & pre-mortem winner
Executionnexload-reasoning-executionDelivery & VerificationAtomic phases, verified test claims