fix(llm-security): misc scanner correctness — reflog FP, diff exact-pass, hex dedupe, SARIF version (#20,#22,#50,#54,#56)

#20 git-forensics' reflog force-push detector had a redundant bare 'reset' term subsuming 'reset:', so any commit subject containing 'reset' tripped it; removed. #22 diff-engine's per-current moved-fallback greedily consumed a baseline candidate a later byte-exact match needed (mislabeling unchanged as new/moved on duplicate fingerprints); a global exact pass now runs before the moved-fallback. #54 memory-poisoning double-reported a 64+ char hex token as both base64 and hex; the base64 check now skips pure-hex tokens.

#56 SARIF output hardcoded driver.version 6.0.0 because the orchestrator called toSARIF() without a version; it now passes the real plugin version from package.json. #50 the VS Code known-malicious blocklist was empty with no explanation (unlike the JetBrains file's 'empty by design' note); added a matching blocklist_note. Suite 2004/0.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Claude-Session: https://claude.ai/code/session_01TcQyMTQfyrsAapaCMPxTtQ

scanners/git-forensics.mjs

@@ -115,7 +115,10 @@ function detectForcePushes(targetPath) {
   try {
     const reflog = git(['reflog', '--format=%H %gD %gs', '-n', '500'], targetPath);
     const lines = reflog.split('\n').filter(Boolean);
-    const resetLines = lines.filter(l => l.includes('reset:') || l.includes('reset'));
+    // v7.8.3 (#20): match only the reflog ACTION `reset:` — a bare `reset`
+    // also matched commit subjects (reflog %gs for a commit is
+    // `commit: <subject>`), so "reset onboarding flow" tripped the detector.
+    const resetLines = lines.filter(l => l.includes('reset:'));
 
     if (resetLines.length > 0) {
       const examples = resetLines.slice(0, 3).map(l => l.slice(0, 80)).join(' | ');

scanners/lib/diff-engine.mjs

@@ -175,7 +175,13 @@ export function diffFindings(baselineFindings, currentFindings) {
     moved: [],
   };
 
-  // Pass 1: Match current findings against baseline
+  // Pass 1a: Global EXACT-match pass (same file, line within threshold).
+  // v7.8.3 (#22): the moved-fallback used to run per-current inside this loop,
+  // so an earlier unmatched current finding greedily consumed a baseline
+  // candidate that a LATER current finding needed for a byte-exact match
+  // (duplicate fingerprints), mislabeling unchanged findings as new/moved.
+  // Exact matches are now claimed globally before any moved-fallback runs.
+  const unmatchedCurrent = [];
   for (const current of currentFindings) {
     const candidates = baselineByFp.get(current.fingerprint);
     if (!candidates) {
@@ -183,7 +189,6 @@ export function diffFindings(baselineFindings, currentFindings) {
       continue;
     }
 
-    // Try exact match first (same file, line within threshold)
     let matched = false;
     for (const baseline of candidates) {
       if (baseline.matched) continue;
@@ -197,9 +202,13 @@ export function diffFindings(baselineFindings, currentFindings) {
         break;
       }
     }
-    if (matched) continue;
+    if (!matched) unmatchedCurrent.push(current);
+  }
 
-    // Try moved match (fingerprint matches, location differs)
+  // Pass 1b: Moved-fallback on the remainder (fingerprint matches, location differs)
+  for (const current of unmatchedCurrent) {
+    const candidates = baselineByFp.get(current.fingerprint);
+    let matched = false;
     for (const baseline of candidates) {
       if (baseline.matched) continue;
       baseline.matched = true;
@@ -211,10 +220,9 @@ export function diffFindings(baselineFindings, currentFindings) {
       matched = true;
       break;
     }
-    if (matched) continue;
 
     // All candidates consumed — this is new
-    results.new.push(current);
+    if (!matched) results.new.push(current);
   }
 
   // Pass 2: Unmatched baseline findings are resolved

scanners/memory-poisoning-scanner.mjs

@@ -326,6 +326,10 @@ function detectEncodedPayloads(content, relPath) {
     BASE64_TOKEN_RE.lastIndex = 0;
     let match;
     while ((match = BASE64_TOKEN_RE.exec(line)) !== null) {
+      // v7.8.3 (#54): a 64+ char pure-hex token (e.g. a sha256) is also
+      // reported by the hex-blob check below — skip the base64 report so one
+      // token yields one finding, labelled hex.
+      if (/^(?:0x)?[0-9a-fA-F]{64,}$/.test(match[0])) continue;
       if (isBase64Like(match[0])) {
         results.push(finding({
           scanner: 'MEM',

scanners/scan-orchestrator.mjs

@@ -283,7 +283,13 @@ async function main() {
   }
 
   // Output: SARIF or JSON, to file (--output-file) or stdout
-  const finalOutput = args.format === 'sarif' ? toSARIF(output) : output;
+  // v7.8.3 (#56): pass the real plugin version from package.json — calling
+  // toSARIF(output) bare hardcoded its stale '6.0.0' default as driver.version.
+  let pluginVersion = '0.0.0';
+  try {
+    pluginVersion = JSON.parse(readFileSync(join(pluginRoot, 'package.json'), 'utf8')).version || pluginVersion;
+  } catch { /* unreadable package.json — fall back to placeholder */ }
+  const finalOutput = args.format === 'sarif' ? toSARIF(output, pluginVersion) : output;
   const jsonStr = JSON.stringify(finalOutput, null, 2) + '\n';
   if (args.outputFile) {
     writeFileSync(args.outputFile, jsonStr);