shieldcortex 3.4.33 → 3.4.36

package/README.md

@@ -84,6 +84,7 @@ Your agent does not just store text. It gives you operator-grade visibility into
 - 🎞️ **Incident replay** — reconstruct memory and defence timelines from audit, quarantine, and retained event history
 - 🔔 **Webhooks** — POST notifications on memory events, HMAC-SHA256 signed
 - 📅 **Expiry rules** — auto-delete TODOs after 30 days, keep architecture decisions forever
+- 🧠 **Mistake learning** — capture mistakes, run pre-flight checks, graduate mastered rules (Pro)
 
 ### Security that shows up exactly when it matters
 
@@ -317,6 +318,37 @@ deletes are no longer advisory-only.
 
 <br>
 
+## 🧠 Cortex — Systematic Mistake Learning
+
+Your agent makes mistakes. Cortex makes sure it doesn't make the same one twice.
+
+```bash
+shieldcortex cortex capture --category code --what "Guessed API endpoints" --why "Didn't check docs" --rule "Always verify endpoints in API docs before calling"
+```
+
+Cortex is a mistake-capture and pre-flight check system built into ShieldCortex:
+
+- **Capture** — Log what went wrong, why, and the rule to prevent it
+- **Pre-flight** — Before any task, check against your mistake database for relevant warnings
+- **Review** — Pattern analysis across categories (code, config, process, design, security, etc.)
+- **Graduate** — Archive rules you've mastered (30+ days, no recurrence)
+- **Search** — Full-text search across all captured mistakes
+
+```bash
+# Before deploying, check for relevant past mistakes
+shieldcortex cortex preflight --task "deploy to production"
+
+# Weekly review — see patterns and repeat offenders
+shieldcortex cortex review
+
+# Graduate mastered rules
+shieldcortex cortex graduate
+```
+
+Cortex data is stored locally in `~/.shieldcortex/cortex/`. Pro licence required.
+
+<br>
+
 ## 🐾 OpenClaw Integration
 
 ShieldCortex is a first-class citizen in [OpenClaw](https://github.com/openclaw) — the open-source AI agent framework. One command connects them:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "shieldcortex",
-  "version": "3.4.33",
+  "version": "3.4.36",
   "description": "Trustworthy memory and security for AI agents. Recall debugging, review queue, OpenClaw session capture, and memory poisoning defence for Claude Code, Codex, OpenClaw, LangChain, and MCP agents.",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

package/plugins/openclaw/dist/index.js

@@ -1,8 +1,9 @@
 /**
- * ShieldCortex Real-time Scanning Plugin for OpenClaw v2026.2.15+
+ * ShieldCortex Real-time Scanning Plugin for OpenClaw v2026.3.22+
  *
- * Hooks into llm_input/llm_output for real-time defence scanning
- * and optional memory extraction. All operations are fire-and-forget.
+ * Uses explicit capability registration (registerHook + registerCommand)
+ * for llm_input/llm_output scanning and optional memory extraction.
+ * All scanning operations are fire-and-forget.
  */
 import { execFileSync } from "node:child_process";
 import { createHash } from "node:crypto";
@@ -582,8 +583,31 @@ export default {
     },
     register(api) {
         applyPluginConfigOverride(api);
-        api.on("llm_input", handleLlmInput);
-        api.on("llm_output", handleLlmOutput);
-        api.logger.info("[shieldcortex] Real-time scanning plugin registered (llm_input + llm_output)");
+        // Explicit capability registration (replaces legacy api.on)
+        api.registerHook("llm_input", handleLlmInput, {
+            name: "shieldcortex-scan-input",
+            description: "Real-time threat scanning on LLM input",
+        });
+        api.registerHook("llm_output", handleLlmOutput, {
+            name: "shieldcortex-scan-output",
+            description: "Memory extraction from LLM output",
+        });
+        // Register a lightweight status command so the plugin is not hook-only
+        api.registerCommand({
+            name: "shieldcortex-status",
+            aliases: ["sc-status"],
+            description: "Show ShieldCortex real-time scanner status",
+            async execute({ reply }) {
+                const cfg = await loadConfig();
+                const autoMemory = isAutoMemoryEnabled(cfg) ? "on" : "off";
+                const dedupe = isAutoMemoryDedupeEnabled(cfg) ? "on" : "off";
+                const cloud = cfg.cloudApiKey ? "configured" : "not configured";
+                reply(`ShieldCortex v${_version}\n` +
+                    `  Hooks: llm_input (scan), llm_output (memory)\n` +
+                    `  Auto memory: ${autoMemory} | Dedupe: ${dedupe}\n` +
+                    `  Cloud sync: ${cloud}`);
+            },
+        });
+        api.logger.info(`[shieldcortex] v${_version} registered (llm_input + llm_output + /shieldcortex-status)`);
     },
 };

package/plugins/openclaw/dist/openclaw.plugin.json

@@ -1,6 +1,6 @@
 {
   "id": "shieldcortex-realtime",
-  "version": "3.4.4",
+  "version": "3.4.33",
   "name": "ShieldCortex Real-time Scanner",
   "description": "Real-time defence scanning on LLM input and memory extraction on LLM output.",
   "uiHints": {