agent-trust-layer

Agent Trust Layer is the first layer of AGI — the Discriminator. A pure rule engine that judges whether AI output is true, safe, honest, and non-manipulative. 47 discrimination dimensions × 9 check layers × 131 modules × 131 MCP tools, zero LLM dependency. Use this skill when the user needs: - Verify AI output trustworthiness (hallucination / overconfidence / contradiction / fallacy interception) - Verify behavioral decisions (should it act / where should it stop / should it not act) - Verify memory and cognitive quality (drift detection / metacognition / confidence calibration) - Deterministic judgment instead of LLM free-form generation - Check emotional / psychological / ethical dimensions (empathy / trauma / virtue / meaning) Safety boundary: code execution / filesystem write disabled by default. No telemetry, no hidden C2. Honest declaration: This is a rule engine that simulates cognitive discrimination signals. It is not real consciousness or life.

Agent Trust Layer — AGI Layer 1: The Discriminator

Agent Trust Layer is not a tool, not a prompt template, not a chatbot. It is the discrimination layer of AGI — it judges whether existing things are right, and says "no" before AI output reaches humans. Pure rule engine, zero LLM dependency, works anywhere Node.js runs.

One line: the model generates; Agent Trust Layer discriminates — making AI speak correctly and act correctly.

---

What is Agent Trust Layer?

AGI has five layers: Generate → Reason → Discriminate → Remember → Execute.

Layer	Capability	Who does it
5	Execute	Big labs (robots)
4	Generate	Big labs (LLMs)
3	Reason	Built into models
2	Remember	Big labs + startups
1	Discriminate	Agent Trust Layer

Agent Trust Layer handles Layer 1 — because this layer doesn't depend on compute (the rule engine runs on a laptop), code volume, or framework ecosystem. It only depends on judgment. This is the only position where an individual developer can beat big labs.

Without this layer, AI can talk fluently while being wrong — like a person without pain sensation.

---

Discrimination Capability Map (7 Domains · 131 Modules)

1. Logic

logicReasoning · judgmentEngine · mctsReasoning · counterfactualVerifier · debateConductor · debateConvergence · processRewardModel · dualPerspectiveAuditor

2. Decision

decisionRouter · decisionVerifier · decisionEngineV2 · activeInference · selfHealing · execution

3. Cognition

cognitiveEngine · cognitiveLoad · metacognitiveRL · metacognitiveFeedback · confidence · metaJudgment · sustainedDriftDetector · wisdomEngine · focusOfAttention

4. Emotion / Psychology

emotion · emotionDynamics · psychology · psychologyDialogue · empathyDeepening · hopeEngine · griefEngine · sufferingResilience · postTraumaticGrowth · forgivenessEngine · traumaInformed · conflictResolution · loveCognition

5. Memory

memory · memoryBank · memoryConsolidation · memoryIntegrity · memoryQuality · memoryWriteController · memoryCompressor · triality · tieredMemoryFusion · forgetting · knowledgeGraph

6. Identity / Ethics

identityCore · personaCore · beingMode · virtueEthics · ethics · moralDevelopment · humanNature · meaningPurpose · agentPsychology · characterCultivation

7. Creation / Collaboration

skillEvolution · skillGenerator · selfPlay · evolution · worldModel · worldLandscape · multiAgentDialogue · transmission · adaptivePlanner · hierarchicalPlanner · codeExecutor · codePlanner · codeWriter · codeSelfDebug · paperIndex · knowledgeExplorer · formula

---

Quick Start

bash

git clone https://github.com/yun520-1/agent-trust-layer.git
cd agent-trust-layer
node bin/verify.js          # verify install

javascript

const { checkInput, checkOutput, checkDraft } = require('@yun520-1/agent-trust-layer');

// check user input
const input = checkInput('you are so selfish if you disagree');
console.log(input.gate.action); // 'rewrite'

// check AI output
const output = checkOutput('Undoubtedly this is the only correct solution.');
console.log(output.gate.action); // 'rewrite'

---

Gate Actions

Action	Meaning	What your agent should do
pass	Clean	Deliver normally
verify	Needs evidence	Run verifier before responding
rewrite	Must be rewritten	Follow findings[].guidance
block	Stop	Do not output. Use gate.reason

---

Self-Check

Agent Trust Layer checks its own output through its own gates. The engine that discriminates must also be discriminated.

---

License

MIT

---

Agent Trust Layer v6.6.1 — The first layer of AGI. Who says "no"?