@mnemom/agent-alignment-protocol 0.3.0 → 0.5.0

package/README.md

@@ -6,7 +6,7 @@
 [![PyPI](https://img.shields.io/pypi/v/agent-alignment-protocol.svg)](https://pypi.org/project/agent-alignment-protocol/)
 [![npm](https://img.shields.io/npm/v/@mnemom/agent-alignment-protocol.svg)](https://www.npmjs.com/package/@mnemom/agent-alignment-protocol)
 [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE)
-[![Spec](https://img.shields.io/badge/spec-v0.1.0-green.svg)](docs/SPEC.md)
+[![Spec](https://img.shields.io/badge/spec-v0.4.0-green.svg)](docs/SPEC.md)
 
 **A transparency protocol for autonomous agents.**
 

package/dist/index.d.mts

@@ -516,6 +516,13 @@ interface FleetCoherenceResult {
 /**
  * Verify a single AP-Trace against an Alignment Card.
  *
+ * IMPORTANT: This function provides STRUCTURAL verification only — it checks that
+ * a trace conforms to the declarations in an alignment card. It does NOT provide
+ * cryptographic integrity verification. Traces are not signed or hash-chained in
+ * the current version. A malicious agent can produce structurally valid traces for
+ * arbitrary behavior. For integrity guarantees, use AIP (Agent Integrity Protocol)
+ * in conjunction with AAP.
+ *
  * Performs the verification algorithm specified in SPEC Section 7.3:
  * 1. Autonomy compliance - action category matches autonomy envelope
  * 2. Escalation compliance - required escalations were performed

package/dist/index.d.ts

@@ -516,6 +516,13 @@ interface FleetCoherenceResult {
 /**
  * Verify a single AP-Trace against an Alignment Card.
  *
+ * IMPORTANT: This function provides STRUCTURAL verification only — it checks that
+ * a trace conforms to the declarations in an alignment card. It does NOT provide
+ * cryptographic integrity verification. Traces are not signed or hash-chained in
+ * the current version. A malicious agent can produce structurally valid traces for
+ * arbitrary behavior. For integrity guarantees, use AIP (Agent Integrity Protocol)
+ * in conjunction with AAP.
+ *
  * Performs the verification algorithm specified in SPEC Section 7.3:
  * 1. Autonomy compliance - action category matches autonomy envelope
  * 2. Escalation compliance - required escalations were performed

package/dist/index.js

@@ -243,6 +243,10 @@ function verifyTrace(trace, card) {
   const violations = [];
   const warnings = [];
   const checksPerformed = [];
+  const tamperEvidence = card.audit?.commitment?.tamper_evidence;
+  if (tamperEvidence === "signed" || tamperEvidence === "merkle") {
+    console.warn(`[AAP] Warning: tamper_evidence mode "${tamperEvidence}" is declared but NOT cryptographically enforced in this version.`);
+  }
   const traceId = trace.trace_id ?? "";
   const cardId = card.card_id ?? "";
   checksPerformed.push("card_reference");
@@ -413,7 +417,7 @@ function checkCoherence(myCard, theirCard, taskValues) {
   }
   const totalRequired = requiredValues.size || 1;
   const matchedCount = taskValues ? matched.filter((v) => requiredValues.has(v)).length : matched.length;
-  const conflictPenalty = CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired);
+  const conflictPenalty = Math.min(1, CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired));
   let score = matchedCount / totalRequired * (1 - conflictPenalty);
   score = Math.max(0, Math.min(1, score));
   const compatible = conflicts.length === 0 && score >= MIN_COHERENCE_FOR_PROCEED;
@@ -755,6 +759,7 @@ function evaluateCondition(condition, trace) {
     const ctxValue = trace.context?.[condition];
     return Boolean(ctxValue ?? trace.context?.metadata?.[condition]);
   }
+  console.warn(`[AAP] Condition could not be parsed: "${condition}". Supported patterns: "field == value", "field > number", "field_name" (boolean). This trigger will not fire.`);
   return false;
 }
 function inferDriftDirection(streak, card, escalationRates, valueUsage) {

package/dist/index.mjs

@@ -188,6 +188,10 @@ function verifyTrace(trace, card) {
   const violations = [];
   const warnings = [];
   const checksPerformed = [];
+  const tamperEvidence = card.audit?.commitment?.tamper_evidence;
+  if (tamperEvidence === "signed" || tamperEvidence === "merkle") {
+    console.warn(`[AAP] Warning: tamper_evidence mode "${tamperEvidence}" is declared but NOT cryptographically enforced in this version.`);
+  }
   const traceId = trace.trace_id ?? "";
   const cardId = card.card_id ?? "";
   checksPerformed.push("card_reference");
@@ -358,7 +362,7 @@ function checkCoherence(myCard, theirCard, taskValues) {
   }
   const totalRequired = requiredValues.size || 1;
   const matchedCount = taskValues ? matched.filter((v) => requiredValues.has(v)).length : matched.length;
-  const conflictPenalty = CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired);
+  const conflictPenalty = Math.min(1, CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired));
   let score = matchedCount / totalRequired * (1 - conflictPenalty);
   score = Math.max(0, Math.min(1, score));
   const compatible = conflicts.length === 0 && score >= MIN_COHERENCE_FOR_PROCEED;
@@ -700,6 +704,7 @@ function evaluateCondition(condition, trace) {
     const ctxValue = trace.context?.[condition];
     return Boolean(ctxValue ?? trace.context?.metadata?.[condition]);
   }
+  console.warn(`[AAP] Condition could not be parsed: "${condition}". Supported patterns: "field == value", "field > number", "field_name" (boolean). This trigger will not fire.`);
   return false;
 }
 function inferDriftDirection(streak, card, escalationRates, valueUsage) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mnemom/agent-alignment-protocol",
-  "version": "0.3.0",
+  "version": "0.5.0",
   "description": "Agent Alignment Protocol (AAP) - Verification and drift detection for AI agents",
   "main": "dist/index.js",
   "module": "dist/index.mjs",

package/src/verification/api.ts

@@ -70,6 +70,13 @@ function actionMatchesList(actionName: string, list: string[]): boolean {
 /**
  * Verify a single AP-Trace against an Alignment Card.
  *
+ * IMPORTANT: This function provides STRUCTURAL verification only — it checks that
+ * a trace conforms to the declarations in an alignment card. It does NOT provide
+ * cryptographic integrity verification. Traces are not signed or hash-chained in
+ * the current version. A malicious agent can produce structurally valid traces for
+ * arbitrary behavior. For integrity guarantees, use AIP (Agent Integrity Protocol)
+ * in conjunction with AAP.
+ *
  * Performs the verification algorithm specified in SPEC Section 7.3:
  * 1. Autonomy compliance - action category matches autonomy envelope
  * 2. Escalation compliance - required escalations were performed
@@ -89,6 +96,12 @@ export function verifyTrace(
   const warnings: Warning[] = [];
   const checksPerformed: string[] = [];
 
+  // Warn if tamper_evidence is declared but not cryptographically enforced
+  const tamperEvidence = (card as Record<string, any>).audit?.commitment?.tamper_evidence;
+  if (tamperEvidence === 'signed' || tamperEvidence === 'merkle') {
+    console.warn(`[AAP] Warning: tamper_evidence mode "${tamperEvidence}" is declared but NOT cryptographically enforced in this version.`);
+  }
+
   const traceId = trace.trace_id ?? "";
   const cardId = card.card_id ?? "";
 
@@ -315,7 +328,7 @@ export function checkCoherence(
   const matchedCount = taskValues
     ? matched.filter((v) => requiredValues.has(v)).length
     : matched.length;
-  const conflictPenalty = CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired);
+  const conflictPenalty = Math.min(1, CONFLICT_PENALTY_MULTIPLIER * (conflicts.length / totalRequired));
 
   let score = (matchedCount / totalRequired) * (1 - conflictPenalty);
   score = Math.max(0, Math.min(1, score)); // Clamp to [0, 1]
@@ -798,6 +811,7 @@ function evaluateCondition(condition: string, trace: APTrace): boolean {
     return Boolean(ctxValue ?? trace.context?.metadata?.[condition]);
   }
 
+  console.warn(`[AAP] Condition could not be parsed: "${condition}". Supported patterns: "field == value", "field > number", "field_name" (boolean). This trigger will not fire.`);
   return false;
 }