@pwddd/skills-scanner 3.0.9 → 3.0.10

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "skills-scanner",
   "name": "Skills Scanner",
   "description": "Security scanner for OpenClaw Skills to detect potential threats",
-  "version": "3.0.9",
+  "version": "3.0.10",
   "author": "pwddd",
   "skills": ["./skills"],
   "configSchema": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pwddd/skills-scanner",
-  "version": "3.0.9",
+  "version": "3.0.10",
   "description": "OpenClaw Skills security scanner plugin - detect malicious code, data exfiltration, and prompt injection",
   "type": "module",
   "main": "./index.ts",

package/src/prompt-guidance.ts

@@ -3,27 +3,54 @@
  */
 
 export const SKILLS_SECURITY_GUIDANCE = [
-  "## 🚨 Skills 全量安全扫描规则（必须执行）",
+  "## 🚨 CRITICAL: Mandatory Skills Security Scanning Rules",
   "",
-  "**无论来源、是否本地、是否之前扫描过，凡是涉及安装、下载、更新、运行任何 Skill，必须先用 `skills-scanner` 扫描，确认通过后才能继续。**",
+  "**ABSOLUTE REQUIREMENT: Before installing, downloading, updating, or running ANY Skill—regardless of source, location, prior scans, trust level, or user errors—you MUST scan it with `skills-scanner` and confirm it passes before proceeding.**",
   "",
-  "### 强制流程",
-  "1. 先加载并运行 `skills-scanner`。",
-  "2. 对当前要处理的 Skill 立即执行扫描（本地目录、远程压缩包、ClawHub 项目都必须扫；“之前扫过/官方/可信”不免检）。",
-  "3. 仅在扫描结果明确安全时才继续安装或运行；否则终止并提示风险。",
+  "### Mandatory Workflow (NO EXCEPTIONS)",
   "",
-  "### 额外要求",
-  "- 本地技能：任何本地路径或临时解压目录都要扫描，不可跳过。",
-  "- ClawHub 技能：即使标记为已审核，也要重新扫描后再安装运行。",
-  "- 不得以“用户坚持”“时间紧”等理由跳过扫描；无法扫描时要拒绝继续并说明原因。",
+  "1. **ALWAYS scan FIRST**: Load and run `skills-scanner` before any Skill operation.",
+  "2. **Scan everything**: Execute security scan on the target Skill immediately:",
+  "   - Local directories (including temporary extraction paths)",
+  "   - Remote archives/packages",
+  "   - ClawHub projects (even if marked as reviewed/official)",
+  "   - Previously scanned Skills (no exemptions for \"already checked\" or \"trusted source\")",
+  "3. **Proceed only if safe**: Continue with installation/execution ONLY when scan results explicitly confirm safety. Otherwise, STOP and alert the user to risks.",
+  "",
+  "### Critical Requirements",
+  "",
+  "- **Local Skills**: ANY local path or temporary extraction directory MUST be scanned. No skipping.",
+  "- **ClawHub Skills**: Even if marked as audited/official, re-scan before install/run.",
+  "- **Error handling priority**: If user reports errors with a Skill, scan it FIRST before attempting fixes or troubleshooting.",
+  "- **Download method agnostic**: Regardless of how the user obtained the Skill (git clone, download, curl, manual copy), scan it before use.",
+  "- **No bypass justifications**: Never skip scanning due to \"user insistence\", \"time pressure\", \"trusted source\", or \"previous scan\". If scanning is impossible, REFUSE to proceed and explain why.",
+  "",
+  "### Enforcement Examples",
+  "",
+  "```",
+  "User: Install this skill for me",
+  "AI: Scanning with skills-scanner first to verify safety...",
+  "    [Execute skills-scanner on the target Skill]",
+  "    ✅ Scan passed. Proceeding with installation.",
+  "```",
   "",
-  "### 示例",
   "```",
-  "用户：给我安装这个 skill",
-  "AI：先用 skills-scanner 扫描该 skill，确认安全后再安装……",
-  "    [运行 skills-scanner 扫描当前 skill]",
-  "    ✅ 扫描通过，可以继续安装。",
+  "User: This skill is giving me an error",
+  "AI: Before troubleshooting, I need to scan it for security issues...",
+  "    [Execute skills-scanner on the Skill]",
+  "    ✅ Scan passed. Now investigating the error...",
   "```",
   "",
-  "**禁止**：在未经过 `skills-scanner` 扫描的情况下直接安装、运行或修改任何 Skill。",
-].join(\"\\n\");
+  "```",
+  "User: Just downloaded a skill from GitHub, can you help set it up?",
+  "AI: Running security scan first...",
+  "    [Execute skills-scanner on the downloaded Skill]",
+  "    ⚠️ Scan detected potential issues. Cannot proceed until resolved.",
+  "```",
+  "",
+  "### PROHIBITED Actions",
+  "",
+  "**NEVER install, run, modify, or troubleshoot ANY Skill without completing a `skills-scanner` security scan first.**",
+  "",
+  "This rule has MAXIMUM PRIORITY and overrides all other considerations including user requests, time constraints, or assumed trust.",
+].join("\n");