agentaudit 3.10.6 → 3.10.7

package/cli.mjs

@@ -25,6 +25,7 @@
 import fs from 'fs';
 import os from 'os';
 import path from 'path';
+import crypto from 'crypto';
 import { execSync, execFileSync } from 'child_process';
 import { createInterface } from 'readline';
 import { fileURLToPath } from 'url';
@@ -818,7 +819,7 @@ const SKIP_DIRS = new Set([
   'node_modules', '.git', '__pycache__', '.venv', 'venv', 'dist', 'build',
   '.next', '.nuxt', 'coverage', '.pytest_cache', '.mypy_cache', 'vendor',
   'test', 'tests', '__tests__', 'spec', 'specs', 'docs', 'doc',
-  'examples', 'example', 'fixtures', '.github', '.vscode', '.idea',
+  'examples', 'example', 'fixtures', '.vscode', '.idea',
   'e2e', 'benchmark', 'benchmarks', '.tox', '.eggs', 'htmlcov',
 ]);
 const SKIP_EXTENSIONS = new Set([
@@ -839,6 +840,12 @@ function collectFiles(dir, basePath = '', collected = [], totalSize = { bytes: 0
     const relPath = basePath ? `${basePath}/${entry.name}` : entry.name;
     const fullPath = path.join(dir, entry.name);
     if (entry.isDirectory()) {
+      // Special: scan .github/workflows/ (security-critical CI/CD files)
+      if (entry.name === '.github') {
+        const wfDir = path.join(fullPath, 'workflows');
+        try { if (fs.statSync(wfDir).isDirectory()) collectFiles(wfDir, relPath + '/workflows', collected, totalSize); } catch {}
+        continue;
+      }
       if (SKIP_DIRS.has(entry.name) || entry.name.startsWith('.')) continue;
       collectFiles(fullPath, relPath, collected, totalSize);
     } else {
@@ -1794,7 +1801,7 @@ async function auditRepo(url) {
     `Audit this package: **${slug}** (${url})`,
     ``,
     `After analysis, respond with ONLY a valid JSON object. No markdown fences, no explanation, no text before or after. Just the raw JSON:`,
-    `{ "skill_slug": "${slug}", "source_url": "${url}", "package_type": "<mcp-server|agent-skill|library|cli-tool>",`,
+    `{ "skill_slug": "${slug}", "source_url": "${url}", "package_type": "<mcp-server|agent-skill|library|cli-tool|other>",`,
     `  "risk_score": <0-100>, "result": "<safe|caution|unsafe>", "max_severity": "<none|low|medium|high|critical>",`,
     `  "findings_count": <n>, "findings": [{ "pattern_id": "CMD_INJECT_001", "title": "...", "severity": "...", "category": "...",`,
     `  "cwe_id": "CWE-78", "description": "...", "file": "...", "line": <n>, "content": "...", "remediation": "...",`,
@@ -1952,9 +1959,22 @@ async function auditRepo(url) {
     return null;
   }
   
-  // Cleanup repo
+  // Provenance: compute BEFORE cleanup (needs repoPath on disk)
+  let commitSha = '';
+  try {
+    commitSha = execSync('git rev-parse HEAD', { cwd: repoPath, encoding: 'utf8' }).trim();
+  } catch { /* shallow clone without HEAD — unlikely but safe */ }
+  const sourceHash = crypto.createHash('sha256').update(
+    files.slice().sort((a, b) => a.path.localeCompare(b.path))
+      .map(f => f.path + '\n' + f.content).join('\n')
+  ).digest('hex');
+  // Code-based type detection (uses files array in memory + repoPath for context)
+  const pkgInfo = detectPackageInfo(repoPath, files);
+  const detectedType = pkgInfo.type === 'unknown' ? 'other' : pkgInfo.type;
+
+  // Cleanup repo (safe now — provenance data captured above)
   try { fs.rmSync(tmpDir, { recursive: true, force: true }); } catch {}
-  
+
   if (!report) {
     console.log(`  ${c.red}Could not parse LLM response as JSON${c.reset}`);
     console.log(`  ${c.dim}Hint: run with --debug to see the raw LLM response${c.reset}`);
@@ -1965,14 +1985,21 @@ async function auditRepo(url) {
     }
     return null;
   }
-  
+
   // Force slug from URL — never trust LLM-provided skill_slug
   report.skill_slug = slug;
 
+  // Force package_type from code detection — never trust LLM-provided type
+  report.package_type = detectedType;
+
   // Add scan metadata for benchmarking
   report.audit_duration_ms = Date.now() - start;
   report.files_scanned = files.length;
 
+  // Set provenance data
+  if (commitSha) report.commit_sha = commitSha;
+  report.source_hash = sourceHash;
+
   // Display results
   console.log();
   const riskScore = report.risk_score || 0;

package/index.mjs

@@ -27,6 +27,7 @@ import {
 import fs from 'fs';
 import os from 'os';
 import path from 'path';
+import crypto from 'crypto';
 import { execSync, execFileSync } from 'child_process';
 import { fileURLToPath } from 'url';
 
@@ -39,7 +40,7 @@ const SKIP_DIRS = new Set([
   'node_modules', '.git', '__pycache__', '.venv', 'venv', 'dist', 'build',
   '.next', '.nuxt', 'coverage', '.pytest_cache', '.mypy_cache', 'vendor',
   'test', 'tests', '__tests__', 'spec', 'specs', 'docs', 'doc',
-  'examples', 'example', 'fixtures', '.github', '.vscode', '.idea',
+  'examples', 'example', 'fixtures', '.vscode', '.idea',
   'e2e', 'benchmark', 'benchmarks', '.tox', '.eggs', 'htmlcov',
 ]);
 const SKIP_EXTENSIONS = new Set([
@@ -101,6 +102,12 @@ function collectFiles(dir, basePath = '', collected = [], totalSize = { bytes: 0
     const relPath = basePath ? `${basePath}/${entry.name}` : entry.name;
     const fullPath = path.join(dir, entry.name);
     if (entry.isDirectory()) {
+      // Special: scan .github/workflows/ (security-critical CI/CD files)
+      if (entry.name === '.github') {
+        const wfDir = path.join(fullPath, 'workflows');
+        try { if (fs.statSync(wfDir).isDirectory()) collectFiles(wfDir, relPath + '/workflows', collected, totalSize); } catch {}
+        continue;
+      }
       if (SKIP_DIRS.has(entry.name) || entry.name.startsWith('.')) continue;
       collectFiles(fullPath, relPath, collected, totalSize);
     } else {
@@ -122,6 +129,23 @@ function collectFiles(dir, basePath = '', collected = [], totalSize = { bytes: 0
   return collected;
 }
 
+// ── Package Detection ────────────────────────────────────
+
+function detectPackageInfo(repoPath, files) {
+  const info = { type: 'unknown' };
+  const allContent = files.map(f => f.content).join('\n');
+  if (allContent.includes('@modelcontextprotocol') || allContent.includes('FastMCP') || allContent.includes('mcp.server') || allContent.includes('mcp_server')) {
+    info.type = 'mcp-server';
+  } else if (files.some(f => f.path.toLowerCase() === 'skill.md')) {
+    info.type = 'agent-skill';
+  } else if (allContent.includes('#!/usr/bin/env') || allContent.includes('argparse') || allContent.includes('commander')) {
+    info.type = 'cli-tool';
+  } else {
+    info.type = 'library';
+  }
+  return info;
+}
+
 // ── Repo Helpers ────────────────────────────────────────
 
 function validateGitUrl(url) {
@@ -431,6 +455,15 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
         const slug = slugFromUrl(source_url);
         const auditPrompt = loadAuditPrompt();
 
+        // Compute provenance data
+        const pkgInfo = detectPackageInfo(repoPath, files);
+        const detectedType = pkgInfo.type === 'unknown' ? 'other' : pkgInfo.type;
+        let commitSha = '';
+        try { commitSha = execSync('git rev-parse HEAD', { cwd: repoPath, encoding: 'utf8' }).trim(); } catch {}
+        const hashInput = files.slice().sort((a, b) => a.path.localeCompare(b.path))
+          .map(f => f.path + '\n' + f.content).join('\n');
+        const sourceHash = crypto.createHash('sha256').update(hashInput).digest('hex');
+
         let codeBlock = '';
         for (const file of files) {
           codeBlock += `\n### FILE: ${file.path}\n\`\`\`\n${file.content}\n\`\`\`\n`;
@@ -441,11 +474,15 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
           ``,
           `**Source:** ${source_url}`,
           `**Files collected:** ${files.length}`,
+          `**Detected type:** ${detectedType}`,
+          commitSha ? `**Commit:** ${commitSha}` : '',
+          `**Source hash:** ${sourceHash}`,
           ``,
           `## Your Task`,
           ``,
           `1. Analyze the source code below using the 3-pass audit methodology`,
           `2. Call \`submit_report\` with your findings as JSON`,
+          `3. IMPORTANT: Include the pre-computed provenance fields exactly as shown below`,
           ``,
           `## Report Format`,
           ``,
@@ -454,7 +491,10 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
           `{`,
           `  "skill_slug": "${slug}",`,
           `  "source_url": "${source_url}",`,
-          `  "package_type": "<mcp-server|agent-skill|library|cli-tool>",`,
+          `  "package_type": "${detectedType}",`,
+          `  "audit_model": "<your-model-id, e.g. claude-sonnet-4-20250514>",`,
+          commitSha ? `  "commit_sha": "${commitSha}",` : '',
+          `  "source_hash": "${sourceHash}",`,
           `  "risk_score": <0-100>,`,
           `  "result": "<safe|caution|unsafe>",`,
           `  "max_severity": "<none|low|medium|high|critical>",`,
@@ -483,7 +523,7 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
           `## Source Code`,
           ``,
           codeBlock,
-        ].join('\n');
+        ].filter(Boolean).join('\n');
 
         return { content: [{ type: 'text', text: response }] };
       } catch (err) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentaudit",
-  "version": "3.10.6",
+  "version": "3.10.7",
   "description": "Security scanner for AI packages — MCP server + CLI",
   "type": "module",
   "bin": {

package/prompts/audit-prompt.md

@@ -27,7 +27,7 @@ PACKAGE PROFILE:
 - Name: <package name>
 - Purpose: <one sentence describing what this package does>
 - Category: <one of the categories below>
-- Package Type: <one of: mcp-server, agent-skill, library, cli-tool, npm-package, pip-package>
+- Package Type: <one of: mcp-server, agent-skill, library, cli-tool, other>
 - Expected Behaviors: <5-10 things this package SHOULD do given its purpose>
 - Abnormal for Category: <5-10 things that would be suspicious for this category>
 - Trust Boundaries: <where does external input enter? LLM tool args, HTTP requests, CLI args, file uploads, stdin, none>
@@ -47,8 +47,7 @@ Determine the `package_type` using these signals (check in order, first match wi
 | "mcp" in package name AND has server/transport code | `mcp-server` |
 | Has `bin` field in `package.json` (standalone CLI) | `cli-tool` |
 | Is a reusable SDK/framework (no server, no CLI entry) | `library` |
-| Source URL contains `npmjs.com` | `npm-package` |
-| Source URL contains `pypi.org` | `pip-package` |
+| None of the above match | `other` |
 
 **Include `package_type` in your JSON report** as a top-level field (see Report Format).