@privacyscrubber/mcp-server 2.3.9 → 2.4.1

package/scrubber-core.cjs

@@ -13,6 +13,8 @@
 
     function getEngine() {
         if (typeof window !== 'undefined' && window.PrivacyScrubberEngine) return window.PrivacyScrubberEngine;
+        if (typeof globalThis !== 'undefined' && globalThis.PrivacyScrubberEngine) return globalThis.PrivacyScrubberEngine;
+        if (typeof self !== 'undefined' && self.PrivacyScrubberEngine) return self.PrivacyScrubberEngine;
         if (typeof global !== 'undefined' && global.PrivacyScrubberEngine) return global.PrivacyScrubberEngine;
         if (typeof require === 'function') {
             try { return require('./ps-pii-engine.cjs'); } catch(e) {}
@@ -37,11 +39,15 @@
 
 function stitchOrphanedNameLines(text, profile) {
     if (!text || !text.includes('\n')) return text;
-    const prof = (profile || 'general').toLowerCase();
+    const prof = (typeof profile === 'string' ? profile : 'general').toLowerCase();
     if (prof === 'medical' || prof === 'general') {
-        text = text.replace(/(Patient\s+Name|Emergency\s+Contact|In\s+Case\s+of\s+Emergency|ICE(?:\s+Contact)?):\s*\r?\n+([A-Z\p{Lu}][A-Za-z0-9&.,'’ \t-]*?[A-Za-z0-9\p{Lu}\p{Ll}])(?=\r?\n|$)/gui, '$1: $2');
+        text = text.replace(/(Patient\s+Name|Emergency\s+Contact|In\s+Case\s+of\s+Emergency|ICE(?:\s+Contact)?|Subject\s+Personnel|Personnel|Employee):\s*\r?\n+([A-Z\p{Lu}][A-Za-z0-9&.,'’ \t-]*?[A-Za-z0-9\p{Lu}\p{Ll}])(?=\r?\n|$)/gui, '$1: $2');
     } else if (prof === 'legal') {
-        text = text.replace(/Defendant:\s*\r?\n+([A-Z][a-zA-Z]+\s[A-Z][a-zA-Z]+)/g, 'Defendant: $1');
+        text = text.replace(/Defendant:\s*\r?\n+([A-Z\p{Lu}][\p{L}'’–-]+\s+[A-Z\p{Lu}][\p{L}'’–-]+)/gu, 'Defendant: $1');
+    } else if (prof === 'personal' || prof === 'vip' || prof === 'executive') {
+        text = text.replace(/(Close\s+Protection\s+Officer|CPO|Bodyguard|Detail\s+Leader|Shift\s+Leader|Protection\s+Agent|Estate\s+Manager|Private\s+Butler|Chauffeur|Personal\s+Driver):\s*\r?\n+([A-Z\p{Lu}][\p{L}'’–-]+\s+[A-Z\p{Lu}][\p{L}'’–-]+)/gu, '$1: $2');
+    } else if (prof === 'agents' || prof === 'ai_agents') {
+        text = text.replace(/(AI\s+Research\s+Scientist|Lead\s+ML\s+Engineer|Machine\s+Learning\s+Engineer|Prompt\s+Engineer|Agent\s+Architect|Agentic\s+Systems\s+Lead|AI\s+Safety\s+Researcher|AI\s+Red\s+Teamer|LLM\s+Security\s+Architect|Agent\s+Ops\s+Engineer|Principal\s+AI\s+Engineer):\s*\r?\n+([A-Z\p{Lu}][\p{L}'’–-]+\s+[A-Z\p{Lu}][\p{L}'’–-]+)/gu, '$1: $2');
     }
     return text;
 }
@@ -59,6 +65,7 @@ function stitchOrphanedNameLines(text, profile) {
  */
 function scrubText(text, customRules = [], tokenLabelMap = {}, profile = 'General', existingSessionMap = {}, isPro = false, ignoreList = null) {
     if (!text) return { scrubbedText: "", tokenMap: {}, count: 0, uniqueUnmasked: new Set(), trialMeta: null, executionMs: 0 };
+    if (typeof profile !== 'string' || !profile) profile = 'General';
     
     const executionStartMs = (typeof performance !== 'undefined' && typeof performance.now === 'function') ? performance.now() : Date.now();
     // Stitch PDF-split name fragments across line breaks (all profiles)
@@ -124,7 +131,7 @@ function scrubText(text, customRules = [], tokenLabelMap = {}, profile = 'Genera
     // Pre-scan input text for pre-existing tokens ([TYPE_N], <TYPE_N>, {{TYPE_N}}, __TYPE_N__)
     // to advance counters and avoid counter collisions or overwriting literal quoted tokens
     if (typeof textToProcess === 'string' && textToProcess) {
-        const existingTokenRegex = /(?:\[|<|\{\{|__)([A-Za-z0-9_]+)_(\d+)(?:\]|>|\}\}|__)/g;
+        const existingTokenRegex = /(?:\[|\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|【|［|<|\{\{?|__)([A-Za-z0-9_]+)_(\d+)(?:\]|\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|】|］|>|\}\}?|__)/g;
         let tMatch;
         while ((tMatch = existingTokenRegex.exec(textToProcess)) !== null) {
             const rawLabel = tMatch[1].toUpperCase();
@@ -377,17 +384,17 @@ function buildRestorationRegexAndRules(tokenMap) {
     const regexParts = [];
 
     sortedKeys.forEach(k => {
-        const inner = k.replace(/^\[|<|\{\{|__|\]|>|\}\}|__/g, '');
+        const inner = k.replace(/^\[|<|\{\{?|__|\]|>|\}\}?|__/g, '');
         const match = inner.match(/^([A-Za-z_0-9]+?)[-_]?(\d+)$/);
         if (match) {
             const label = match[1];
             const baseIndex = parseInt(match[2], 10);
             const aliases = getLabelAliases(label);
             
-            const escapedAliases = aliases.map(a => a.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'));
+            const escapedAliases = aliases.map(a => a.replace(/[.*+?^${}()|[\]\\]/g, '\\$&').replace(/[_-]/g, '\\\\?[_-]'));
             const aliasesGroup = `(?:${escapedAliases.join('|')})`;
             
-            const looseRegex = '(?:\\[?\\s*' + aliasesGroup + '[-_\\s]*0*' + baseIndex + '\\s*\\]?|<\\s*' + aliasesGroup + '[-_\\s]*0*' + baseIndex + '\\s*>|\\{\\{\\s*' + aliasesGroup + '[-_\\s]*0*' + baseIndex + '\\s*\\}\\}|__\\s*' + aliasesGroup + '[-_\\s]*0*' + baseIndex + '\\s*__)(?:\'s|’s|s|[а-яёА-ЯЁ]{1,3})?';
+            const looseRegex = '(?:(?:\\[|\\\\u005[Bb]|%5[Bb]|&(?:amp;)?#0*91;|&(?:amp;)?#[xX]0*5[Bb];|&(?:amp;)?lbrack;|【|［|〔|「|『)?\\s*[*`~]*' + aliasesGroup + '[-_\\\\\\s]*0*' + baseIndex + '[*`~]*[.,;:!?]?\\s*(?:\\]|\\\\u005[Dd]|%5[Dd]|&(?:amp;)?#0*93;|&(?:amp;)?#[xX]0*5[Dd];|&(?:amp;)?rbrack;|】|］|〕|」|』)?|<\\s*' + aliasesGroup + '[-_\\\\\\s]*0*' + baseIndex + '[.,;:!?]?\\s*>|\\{\\{?\\s*' + aliasesGroup + '[-_\\\\\\s]*0*' + baseIndex + '[.,;:!?]?\\s*\\}\\}?|__\\s*' + aliasesGroup + '[-_\\\\\\s]*0*' + baseIndex + '\\s*__)(?:\'s|’s|s|[а-яёА-ЯЁ]{1,3})?';
             looseRules.push({ patternStr: looseRegex, regex: new RegExp('^' + looseRegex + '$', 'i'), originalKey: k });
             regexParts.push(looseRegex);
         } else {
@@ -437,7 +444,14 @@ function cleanAIPromptPrefix(text) {
     cleaned = cleaned.replace(/^(?:Here (?:is|are) (?:the )?(?:redacted|scrubbed|sanitized|processed|clean|updated|modified) (?:text|output|version|data).*?[:\n]+|\*\*Scrubbed Text\*\*[:\n]+|### Scrubbed Text[:\n]+)/i, '');
     cleaned = cleaned.replace(/^\s*Edit\s*\n+/i, "");
     cleaned = cleaned.replace(/\s*\bEdit\s+in\s+a\s+page\b\s*$/i, "");
-    // 3. Strip stray leading colons, semicolons, or separators left by stripped icons/artifact headers
+    // 3. Strip echoed system instructions and PrivacyScrubber / ZTDS pseudonymization disclaimers
+    cleaned = cleaned.replace(/^\s*(?:Understood|Acknowledged|Got it|Sure|OK|Okay|Note taken|Understood[.,!])[^\n]*\n+/i, "");
+    cleaned = cleaned.replace(/(?:^|\n*)\s*(?:----------------------\s*\n*)?\[(?:ZTDS(?:™|\(TM\))?\s*Protocol|Privacy Note|Context: identifiers|Privacy Scrubber Mode|SYSTEM INSTRUCTION: DATA PRIVACY MODE)[\s\S]*?(?:in your response\.?\]|(?:\]\s*(?:\n*----------------------\s*)?))(?:\n*|$)/gi, "\n\n");
+    cleaned = cleaned.replace(/(?:^|\n*)\s*(?:----------------------\s*\n*)?\[(?:ZTDS(?:™|\(TM\))?\s*Protocol|Privacy Note|Context: identifiers|Privacy Scrubber Mode|SYSTEM INSTRUCTION: DATA PRIVACY MODE|Secured by PrivacyScrubber)[^\n\]]*\]\s*(?:\n*----------------------\s*)?(?:\n*|$)/gi, "\n\n");
+    cleaned = cleaned.replace(/(?:^|\n*)\s*(?:Direct identifiers in this prompt were (?:pseudonymized|client-side sanitized)|I will retain all bracketed tokens)[^\n]*(?:\n*|$)/gi, "\n\n");
+    // 4. Strip stray leading colons, semicolons, or separators left by stripped icons/artifact headers
+    cleaned = cleaned.replace(/^(?:\s*----------------------\s*\n+)+/, "");
+    cleaned = cleaned.replace(/(?:\n+\s*----------------------\s*)+$/, "");
     cleaned = cleaned.replace(/^[:;|\-\—\–]+(?=\n|$)/, "");
     cleaned = cleaned.replace(/^[:;]+\s*/, "");
     return cleaned.trim();
@@ -1554,27 +1568,17 @@ function hydrateRegex(rule) {
                 if (isCurrentActive) item.classList.add('active');
                 
                 let labelText = p.label;
-                item.innerText = labelText;
-
-                if (!isPro && p.id !== 'general') {
-                    item.style.opacity = '0.5';
-                    item.style.cursor = 'not-allowed';
-                    item.title = 'PRO Feature';
-                    item.innerText = labelText + ' 🔒';
-                } else if (isCurrentActive) {
+                if (isCurrentActive) {
                     item.innerText = labelText + ' ✓';
+                } else if (!isPro && p.id !== 'general') {
+                    item.innerText = labelText + ' (5k Trial)';
+                    item.title = 'Free 5,000-character trial per scrub';
+                } else {
+                    item.innerText = labelText;
                 }
 
                 item.addEventListener('click', (ev) => {
                     ev.stopPropagation();
-                    if (!isPro && p.id !== 'general') {
-                        if (typeof showInPageToast === 'function') {
-                            showInPageToast('Specialized Profiles require PRO upgrade.', 'warning');
-                        }
-                        menu.remove();
-                        return;
-                    }
-
                     chrome.storage.local.set({ ps_active_profile: p.id }, () => {
                         if (typeof showInPageToast === 'function') {
                             showInPageToast(`Profile Switched to ${p.label}`, 'success');
@@ -1644,7 +1648,7 @@ function hydrateRegex(rule) {
             if (window.showInPageToast) window.showInPageToast("✓ Session exported successfully", "success");
         } catch (e) {
             console.error("PrivacyScrubber Export Error:", e);
-            if (window.showInPageToast) window.showInPageToast("❌ Export failed.", "error");
+            if (window.showInPageToast) window.showInPageToast("Export failed.", "error");
         }
     }
 
@@ -1771,6 +1775,8 @@ function hydrateRegex(rule) {
     }
     if (typeof globalThis !== 'undefined') {
         globalThis.PrivacyScrubberCore = coreInstance;
+        globalThis.scrubText = scrubText;
+        globalThis.unscrubText = coreInstance.unscrubText;
     }
     if (typeof module !== 'undefined' && module.exports) {
         module.exports = {

package/server.json

@@ -1,27 +1,50 @@
 {
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
   "name": "io.github.moxno/privacyscrubber-mcp",
+  "title": "PrivacyScrubber PII Masking MCP",
   "description": "Zero-Trust PII and secrets sanitizer for Cursor, Claude, and AI agents. 100% in-memory.",
+  "websiteUrl": "https://privacyscrubber.com/pii-mcp/?utm_source=mcp_registry&utm_medium=server_json&utm_campaign=mcp_docs",
   "repository": {
     "url": "https://github.com/moxno/privacyscrubber-mcp",
     "source": "github"
   },
-  "version": "2.3.9",
+  "version": "2.4.1",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@privacyscrubber/mcp-server",
-      "version": "2.3.9",
+      "version": "2.4.1",
+      "registryBaseUrl": "https://registry.npmjs.org",
+      "runtimeHint": "npx",
       "transport": {
         "type": "stdio"
       },
       "environmentVariables": [
         {
+          "name": "PRIVACYSCRUBBER_KEY",
           "description": "Optional PRO or TEAMS license key to unlock specialized detection profiles.",
           "isRequired": false,
           "format": "string",
-          "isSecret": true,
-          "name": "PRIVACYSCRUBBER_KEY"
+          "isSecret": true
+        }
+      ]
+    },
+    {
+      "registryType": "pypi",
+      "identifier": "privacyscrubber-mcp",
+      "version": "2.4.1",
+      "registryBaseUrl": "https://pypi.org",
+      "runtimeHint": "uvx",
+      "transport": {
+        "type": "stdio"
+      },
+      "environmentVariables": [
+        {
+          "name": "PRIVACYSCRUBBER_KEY",
+          "description": "Optional PRO or TEAMS license key to unlock specialized detection profiles.",
+          "isRequired": false,
+          "format": "string",
+          "isSecret": true
         }
       ]
     }