@blamejs/exceptd-skills 0.13.57 → 0.13.59

package/CHANGELOG.md

@@ -1,5 +1,40 @@
 # Changelog
 
+## 0.13.59 — 2026-05-22
+
+Air-gap mode honored by `--upstream-check` and the `collect` envelope. `doctor` subchecks surface freshness timestamps + walk-cap markers. `--collectors` text matches its JSON.
+
+### Bugs
+
+- **`run --upstream-check --air-gap` was making the registry call anyway.** The upstream-check helper had no air-gap awareness, and the run path didn't gate the call. The refusal now lives at the central upstream-check dispatch so any future caller inherits it; the result envelope carries `upstream_check.air_gap_blocked: true` and `source: "air-gap"` so consumers see the refusal happened.
+- **`doctor --ai-config` walked unbounded** — 48k+ entries under `~/.claude/` (conversation logs, cache, plugin tarballs) before finishing. The walk now caps at 4 depth + 5000 files and skips known-noisy subdir names (`node_modules`, `.git`, `.cache`, `logs`, `sessions`, `conversations`, `history`, `tmp`, `cache`). When the cap fires, `walk_truncated: true` and `walk_caps: { max_depth, max_files }` surface so operators see the bound.
+- **`doctor --ai-config` text mode still said "manual ACL review noted for any sensitive files found"** on Windows even though the runtime ACL audit lands real findings via `icacls`. The placeholder dated to the original POSIX-only implementation; replaced with a description that matches the actual check.
+- **`doctor --collectors` text mode was a strict subset of JSON** — the count of policy-skipped playbooks was visible but the names were `--json`-only. Text now enumerates the first 5 names + a `+N more` indicator so terminal operators see the same actionable information.
+
+### Features
+
+- **`doctor --currency`** surfaces `oldest_last_threat_review`, `newest_last_threat_review`, `max_days_since_review`, and `checked_at` so operators can answer "is my skill catalog stale?" without parsing the per-skill report.
+- **`doctor --rfcs`** surfaces `index_last_modified` + `index_age_days` from the RFC index mtime so operators can answer "is the offline RFC catalog fresh?" without running a separate refresh.
+- **`collect` envelope** surfaces `air_gap_mode` at the top of the result so downstream `run --evidence -` and AI consumers see the collection-time mode propagating. Collectors themselves don't currently make network calls; the marker flags the collection context for future collector additions.
+
+## 0.13.58 — 2026-05-22
+
+Air-gap mode now blocks every refresh source (not just GHSA/OSV). `doctor` help text catches up with runtime flag set, surfaces a fix status when nothing to remediate, and refuses unknown flags. `ask` stops returning false-positive substring matches and learns identity / phishing / SSO / famous-attack vocabulary.
+
+### Bugs
+
+- **`refresh --air-gap` was a partial guarantee.** GHSA + OSV honored `ctx.airGap` at the source-module level. `kev`, `epss`, `nvd`, `rfc`, and `pins` fell through to their live-network branches when neither `--from-fixture` nor `--from-cache` was wired up — operators got a banner promising no egress while CISA KEV updates were being applied to the catalog. The air-gap guard now lives at the central source-dispatch (`runOne` in `lib/refresh-external.js`) so the guarantee holds uniformly across every source. Sources with a `fixtures.<name>` or `cacheDir` offline path still run normally.
+- **`exceptd doctor --help` advertised 6 flags but the runtime accepted 10.** `--collectors`, `--ai-config`, `--exit-codes`, and `--shipped-tarball` were dispatchable but undiscoverable. The help block now enumerates the full set, with one-paragraph descriptions of what each surface checks. `--exit-codes` is now operator-discoverable as the canonical EXIT_CODES contract dump.
+- **`doctor --cves` breakdown arithmetic** silently dropped `BUG-*` and any non-CVE/non-MAL prefix entries from the named totals. `312 entries (301 CVE + 8 MAL)` is now `312 entries (3 BUG + 301 CVE + 8 MAL)` — exact sum. Output adds a structured `by_prefix` map so the breakdown is data-driven against the live catalog instead of hardcoded against two prefixes (same regression class as the v0.13.6 MAL fix, now generalized).
+- **`doctor --fix` silently no-op'd when nothing was wrong.** Operators couldn't distinguish "we tried and were already healthy" from "we tried and failed silently." The summary now carries `fix_status: "already_present"` + `fix_skipped_reason` on a healthy install. Existing `fix_applied` / `fix_attempted` / `fix_partial` states unchanged.
+- **`doctor` accepted unknown flags as no-ops** instead of refusing. `doctor --singatures` (typo), `doctor --data` (renamed-out flag), `doctor --bogus-flag-xyz` all ran a full default scan and exited 0. Refusal is now structured: ok:false + `unknown_flags[]` with did-you-mean candidates + `known_flags[]` listing every accepted flag. Non-zero exit.
+- **`ask` substring matching produced stopword false positives.** `"the"` substring-hit `"authentication"` in the haystack, so any query containing a stopword inflated scores. The pure-stopword query `"the the the the"` routed confidently to ai-api. Fix: tokenize the haystack with the same splitter as the question and match by whole-token Set membership instead of `String.includes`. Plus a stopword filter applied after synonym expansion to drop common English words that would otherwise dilute scoring.
+
+### Features
+
+- **`run` / `ai-run` / `ci` envelopes surface `air_gap_mode` at the top.** Stdout-parsing consumers can detect that air-gap was active without descending into `phases.govern`. Mirrors the existing result-hoist pattern (`verdict`, `rwep_score`, `evidence_completeness`, `attestation_path`).
+- **`ask` synonym map grew to cover identity / phishing / SSO / BEC / famous-attack vocabulary.** New keys: `phish*`, `sso`, `oauth`, `saml`, `okta`, `entra`, `bec`, `deepfake`, `left-pad`, `event-stream`, `shai-hulud`, `agentic`, `rogue`, `credential theft`, `developer laptop`. Queries like `"I think we got phished"` now top-route to identity-sso-compromise; `"left-pad style attack"` to library-author; `"credential theft from developer laptop"` to secrets (with cred-stores in top 3).
+
 ## 0.13.57 — 2026-05-22
 
 `attest list` populates the signed field. README + verb help document the ai-run JSONL event grammar. `--block-on-jurisdiction-clock` clarifies pending-vs-started semantics.

package/bin/exceptd.js

@@ -1989,14 +1989,42 @@ Subchecks:
                           Currently scoped to: regenerate the local Ed25519
                           private key when keys/public.pem exists but
                           .keys/private.pem is absent. Does NOT modify any
-                          file outside .keys/.
-  (no flag)               All four subchecks above (sans --registry-check
-                          unless explicitly requested), plus signing-status
-                          (private key presence under .keys/).
+                          file outside .keys/. When the key is already
+                          present, --fix is a no-op (surfaces fix_status:
+                          "already_present" so callers can distinguish
+                          clean-state from broken-state).
+  --collectors            Audit the per-playbook collector layer:
+                          which playbooks have a collector module under
+                          lib/collectors/, which are policy-skipped by
+                          design, and which collectors haven't been wired
+                          up yet. JSON emits has_collector / policy_skips /
+                          without_collector arrays.
+  --ai-config             Walk the operator's AI-assistant configuration
+                          files (~/.claude/, ~/.cursor/, ~/.codeium/,
+                          ~/.aider/, ~/.continue/) and surface sensitive
+                          content (API keys, tokens, MCP server
+                          definitions) plus on Windows the icacls ACL
+                          state. Combine with --fix to harden ACLs.
+                          Opt-in — never part of the default scan.
+  --exit-codes            Dump the canonical EXIT_CODES table as JSON.
+                          Useful for CI / scripting consumers that want
+                          the documented exit-code contract without parsing
+                          help text. Off by default.
+  --shipped-tarball       Round-trip the verify-shipped-tarball gate:
+                          npm pack → extract to a tempdir → run
+                          lib/verify.js against the extracted tree.
+                          Surfaces the signature-regression class where
+                          source-tree verify passes but the published
+                          tarball fails. Off by default.
+  (no flag)               --signatures + --currency + --cves + --rfcs +
+                          signing-status. --registry-check, --collectors,
+                          --ai-config, --exit-codes, --shipped-tarball are
+                          opt-in and never run as part of the default scan.
 
 Flags:
   --json                  Emit JSON (default is human-readable text).
   --pretty                Indented JSON output (implies --json).
+  --air-gap               Suppress the --registry-check network probe.
 
 Output: checks{} per subcheck + summary{all_green, issues_count}.`,
     "ai-run": `ai-run <playbook> — streaming JSONL contract for AI-driven runs (v0.11.0).
@@ -2326,7 +2354,23 @@ function cmdCollect(runner, args, runOpts, pretty) {
   // `exceptd run <playbook> --evidence -` to drive a real verdict.
   // Human-rendered version is concise so an interactive operator can
   // see what the collector found without parsing the JSON.
-  emit({ verb: "collect", playbook_id: playbookId, ...submission }, pretty, (obj) => {
+  // Audit 3 A.4: surface air_gap_mode on the collect envelope so the
+  // downstream `run --evidence -` sees the mode propagating from the
+  // collection step. Collectors themselves currently make no network
+  // calls — but the flag's intent is to flag the collection context for
+  // any future collector that might.
+  // Also honor _meta.air_gap_mode on the playbook itself — playbooks
+  // like secrets / cred-stores / containers declare air-gap intrinsically
+  // and `run` honors that even without --air-gap. Collect must mirror so
+  // automation downstream sees the same intrinsic mode.
+  let pbMetaAirGap = false;
+  try { pbMetaAirGap = !!(runner.loadPlaybook(playbookId)?._meta?.air_gap_mode); }
+  catch { /* playbook load shouldn't fail here — collector exists — but be defensive */ }
+  const collectAirGap = !!(runOpts.airGap || process.env.EXCEPTD_AIR_GAP === "1" || pbMetaAirGap);
+  // Spread `submission` first, then explicit fields, so a submission key
+  // named `air_gap_mode` (currently always undefined but defensive against
+  // future collector contracts) can't clobber the envelope marker.
+  emit({ verb: "collect", playbook_id: playbookId, ...submission, air_gap_mode: collectAirGap }, pretty, (obj) => {
     const lines = [];
     const meta = obj.collector_meta || {};
     lines.push(`collect: ${obj.playbook_id}  (${meta.collector_version || "?"} on ${meta.platform || "?"})`);
@@ -3146,25 +3190,47 @@ function cmdRun(runner, args, runOpts, pretty) {
   // Network bounded by an 8s timeout; degrades gracefully when offline.
   let upstreamCheck = null;
   if (args["upstream-check"]) {
-    try {
-      const cliPath = path.join(PKG_ROOT, "lib", "upstream-check-cli.js");
-      const res = spawnSync(process.execPath, [cliPath, "--timeout", "5000"], {
-        encoding: "utf8",
-        cwd: PKG_ROOT,
-        timeout: 8000,
-      });
-      try { upstreamCheck = JSON.parse((res.stdout || "").trim()); } catch { /* fall through */ }
-      if (upstreamCheck && upstreamCheck.behind) {
-        process.stderr.write(`[exceptd run --upstream-check] STALE: local v${upstreamCheck.local_version} < published v${upstreamCheck.latest_version} (published ${upstreamCheck.latest_published_at}, ${upstreamCheck.days_since_latest_publish}d ago). Continuing with local catalog. Run \`npm update -g @blamejs/exceptd-skills\` or \`exceptd refresh --network\` to consume the latest.\n`);
+    // Audit 3 A.6: --air-gap must refuse the registry probe. The
+    // upstream-check helper has no air-gap awareness of its own; the
+    // central refusal lives here so any future caller of --upstream-check
+    // inherits it.
+    if (runOpts.airGap || process.env.EXCEPTD_AIR_GAP === "1") {
+      upstreamCheck = {
+        ok: false,
+        source: "air-gap",
+        air_gap_blocked: true,
+        skipped_reason: "--upstream-check would query the npm registry; refused under --air-gap.",
+      };
+    } else {
+      try {
+        const cliPath = path.join(PKG_ROOT, "lib", "upstream-check-cli.js");
+        const res = spawnSync(process.execPath, [cliPath, "--timeout", "5000"], {
+          encoding: "utf8",
+          cwd: PKG_ROOT,
+          timeout: 8000,
+        });
+        try { upstreamCheck = JSON.parse((res.stdout || "").trim()); } catch { /* fall through */ }
+        if (upstreamCheck && upstreamCheck.behind) {
+          process.stderr.write(`[exceptd run --upstream-check] STALE: local v${upstreamCheck.local_version} < published v${upstreamCheck.latest_version} (published ${upstreamCheck.latest_published_at}, ${upstreamCheck.days_since_latest_publish}d ago). Continuing with local catalog. Run \`npm update -g @blamejs/exceptd-skills\` or \`exceptd refresh --network\` to consume the latest.\n`);
+        }
+      } catch (e) {
+        upstreamCheck = { ok: false, error: e.message, source: "offline" };
       }
-    } catch (e) {
-      upstreamCheck = { ok: false, error: e.message, source: "offline" };
     }
   }
 
   const result = runner.run(playbookId, directiveId, submission, runOpts);
   if (result && upstreamCheck) result.upstream_check = upstreamCheck;
 
+  // Audit 3 A.2: surface air_gap_mode at the top of the envelope so
+  // stdout-parsing consumers can detect that the run honored --air-gap
+  // (or the playbook's _meta.air_gap_mode flag) without descending into
+  // phases.govern. Mirrors the run-result hoist pattern (verdict,
+  // rwep_score, evidence_completeness, attestation_path).
+  if (result) {
+    result.air_gap_mode = !!(pb._meta?.air_gap_mode || runOpts.airGap);
+  }
+
   // v0.11.9 (#113/#114): surface --operator and --ack in the run result so
   // operators see the attribution + consent state without inspecting the
   // attestation file. Pre-0.11.9 these were persisted to disk only.
@@ -5920,6 +5986,27 @@ function cmdDoctor(runner, args, runOpts, pretty) {
   const wantJson = !!args.json || !!args.pretty;
   const indent = !!args.pretty;
 
+  // Refuse unknown flags rather than silently running the full default
+  // scan. Pre-fix, typos and renamed-out flags all returned exit 0 with
+  // the operator believing they'd run a targeted check.
+  const KNOWN_DOCTOR_FLAGS = new Set([
+    "json", "pretty", "fix", "air-gap",
+    "signatures", "currency", "cves", "rfcs", "registry-check",
+    "ai-config", "collectors", "exit-codes", "shipped-tarball",
+    // Global flags the parser may inject regardless of verb.
+    "_", "json-stdout-only", "_jsonMode",
+  ]);
+  const unknownFlags = Object.keys(args).filter(k => !KNOWN_DOCTOR_FLAGS.has(k));
+  if (unknownFlags.length > 0) {
+    const dym = unknownFlags.map(f => {
+      const candidates = [...KNOWN_DOCTOR_FLAGS].filter(k => typeof k === "string" && k.length >= 2 && (k.includes(f) || f.includes(k) || levenshtein1(f, k) <= 1));
+      return { flag: `--${f}`, did_you_mean: candidates.slice(0, 3).map(c => `--${c}`) };
+    });
+    return emitError(`doctor: unknown flag(s): ${unknownFlags.map(f => `--${f}`).join(", ")}`,
+      { verb: "doctor", unknown_flags: dym, known_flags: [...KNOWN_DOCTOR_FLAGS].filter(k => k !== "_" && !k.startsWith("_") && k !== "json-stdout-only").sort().map(k => `--${k}`) },
+      pretty);
+  }
+
   // `doctor --exit-codes` dumps the canonical exit-code table as JSON so
   // operator-facing docs cannot drift from runtime behavior. Short-circuit
   // before the regular health checks since the dump is informational.
@@ -6039,12 +6126,25 @@ function cmdDoctor(runner, args, runOpts, pretty) {
         const stale = parsed.currency_report.filter(s => s.action_required || s.currency_label !== "current");
         const critical = parsed.currency_report.filter(s => s.currency_score !== undefined && s.currency_score < 50);
         const ok = stale.length === 0 && !parsed.action_required;
+        // Audit 3 B.7: surface the freshest + stalest last_threat_review
+        // so operators can answer "is my data stale?" without parsing the
+        // full report. Falls back gracefully when the upstream report
+        // omits the per-skill date.
+        const dates = parsed.currency_report
+          .map(s => s.last_threat_review)
+          .filter(d => typeof d === "string" && /^\d{4}-\d{2}-\d{2}$/.test(d))
+          .sort();
+        const minDaysSince = dates.length ? Math.floor((Date.now() - new Date(dates[0]).getTime()) / 86400000) : null;
         checks.currency = {
           ok,
           total_skills: parsed.currency_report.length,
           stale_skills: stale.map(s => s.skill),
           critical_stale: critical.map(s => s.skill),
           critical_count: parsed.critical_count || 0,
+          oldest_last_threat_review: dates[0] || null,
+          newest_last_threat_review: dates[dates.length - 1] || null,
+          max_days_since_review: minDaysSince,
+          checked_at: new Date().toISOString(),
         };
         if (!ok) issues.push("currency");
       } else {
@@ -6079,22 +6179,37 @@ function cmdDoctor(runner, args, runOpts, pretty) {
       // MAL-* (malicious package) entries silently dropped from the
       // doctor report — operators reading "34 entries" assumed the
       // Shai-Hulud / TanStack worm intel had been removed when it was
-      // present all along. Read the catalog and report both totals.
+      // present all along. Read the catalog and report all prefix
+      // groups — the previous CVE+MAL-only enumeration silently dropped
+      // any other prefix (BUG-*, SNYK-*, etc.) from the breakdown even
+      // though they were summed into total. Same regression class as
+      // the original CVE-only fix; this generalizes it.
       let total = null;
       let cve_count = null;
       let mal_count = null;
+      let by_prefix = null;
       try {
         const catalog = require(path.join(PKG_ROOT, "data", "cve-catalog.json"));
         const keys = Object.keys(catalog).filter((k) => !k.startsWith("_"));
         cve_count = keys.filter((k) => k.startsWith("CVE-")).length;
         mal_count = keys.filter((k) => k.startsWith("MAL-")).length;
         total = keys.length;
+        // Enumerate every prefix present so future additions (BUG-*,
+        // SNYK-*, GHSA-*, RUSTSEC-*) surface in the breakdown instead
+        // of vanishing into the total - sum-of-named-prefixes gap.
+        by_prefix = {};
+        for (const k of keys) {
+          const m = k.match(/^([A-Z]+)-/);
+          const p = m ? m[1] : "OTHER";
+          by_prefix[p] = (by_prefix[p] || 0) + 1;
+        }
       } catch { /* fall through with nulls */ }
       checks.cves = {
         ok,
         total,
         cve_count,
         mal_count,
+        by_prefix,
         drift: driftMatch ? Number(driftMatch[1]) : 0,
         ...(ok ? {} : { exit_code: res.status, raw: text.slice(0, 500) }),
       };
@@ -6117,10 +6232,23 @@ function cmdDoctor(runner, args, runOpts, pretty) {
       const rfcRows = (text.match(/^RFC-\d+/gm) || []).length;
       const driftMatch = text.match(/drift[:\s]+(\d+)/i);
       const ok = res.status === 0;
+      // Audit 3 B.7: surface the RFC catalog file's mtime + age so
+      // operators can answer "is the offline RFC index fresh?" without
+      // running a separate refresh.
+      let rfcMtime = null;
+      let rfcAgeDays = null;
+      try {
+        const rfcPath = path.join(PKG_ROOT, "data", "rfc-index.json");
+        const st = fs.statSync(rfcPath);
+        rfcMtime = st.mtime.toISOString();
+        rfcAgeDays = Math.floor((Date.now() - st.mtimeMs) / 86400000);
+      } catch { /* file may not exist on contributor checkouts */ }
       checks.rfcs = {
         ok,
         total: rfcRows,
         drift: driftMatch ? Number(driftMatch[1]) : 0,
+        index_last_modified: rfcMtime,
+        index_age_days: rfcAgeDays,
         ...(ok ? {} : { exit_code: res.status, raw: text.slice(0, 500) }),
       };
       if (!ok) issues.push("rfcs");
@@ -6271,18 +6399,47 @@ function cmdDoctor(runner, args, runOpts, pretty) {
     const findings = [];
     let scannedDirs = 0;
     let scannedFiles = 0;
-    function walk(absDir, displayRoot, rel) {
+    let walkAborted = false;
+    // Audit 3 B.9: cap depth + file count to bound the walk. Without
+    // these, doctor --ai-config can read 48k+ entries under ~/.claude
+    // (conversation logs, cache dirs, plugin tarballs) before finishing.
+    // The five SENSITIVE_PATTERNS files all live within ~3 levels of an
+    // AI-assistant config root, so 4 is generous. The skipped dirs are
+    // known non-config noise that shouldn't carry credentials.
+    const MAX_DEPTH = 4;
+    const MAX_FILES = 5000;
+    const SKIP_DIR_NAMES = new Set([
+      'node_modules', '.git', '.cache', 'logs', 'log',
+      'sessions', 'session', 'transcripts', 'transcript',
+      'conversations', 'history', 'tmp', 'temp', 'cache',
+    ]);
+    function walk(absDir, displayRoot, rel, depth = 0) {
+      if (walkAborted) return;
+      if (depth > MAX_DEPTH) return;
+      if (scannedFiles > MAX_FILES) {
+        walkAborted = true;
+        return;
+      }
       if (!fs.existsSync(absDir)) return;
       let entries;
       try { entries = fs.readdirSync(absDir, { withFileTypes: true }); }
       catch { return; }
       for (const e of entries) {
+        if (walkAborted) return;
+        if (e.isDirectory() && SKIP_DIR_NAMES.has(e.name.toLowerCase())) continue;
         const childAbs = path.join(absDir, e.name);
         const childRel = rel ? rel + '/' + e.name : e.name;
         if (e.isDirectory()) {
-          walk(childAbs, displayRoot, childRel);
+          walk(childAbs, displayRoot, childRel, depth + 1);
         } else if (e.isFile()) {
           scannedFiles++;
+          // Check the file cap immediately after the increment so a
+          // single large directory doesn't process tens of thousands of
+          // entries before the next recursive call catches the bound.
+          if (scannedFiles > MAX_FILES) {
+            walkAborted = true;
+            return;
+          }
           if (!SENSITIVE_PATTERNS.some((re) => re.test(e.name))) continue;
           let st;
           try { st = fs.statSync(childAbs); } catch { continue; }
@@ -6375,6 +6532,8 @@ function cmdDoctor(runner, args, runOpts, pretty) {
       severity: errorFindings.length > 0 && fixesFailed > 0 ? 'warn' : (errorFindings.length > 0 && !args.fix ? 'warn' : 'info'),
       scanned_dirs: scannedDirs,
       scanned_files: scannedFiles,
+      walk_truncated: walkAborted,
+      walk_caps: { max_depth: MAX_DEPTH, max_files: MAX_FILES },
       directories_inspected: AI_CONFIG_DIRS.map((d) => d.display),
       sensitive_patterns: ['settings.json', 'mcp.json', '*.mcp_config.json', 'api_key*', '*.token', '*.credentials'],
       findings,
@@ -6543,6 +6702,23 @@ function cmdDoctor(runner, args, runOpts, pretty) {
     }
   }
 
+  // Audit 3 B.3: --fix was passed but nothing to fix. Pre-fix this was
+  // silently a no-op — operators couldn't distinguish "we tried and were
+  // already healthy" from "we tried and failed silently." Now surfaces a
+  // structured fix_status so callers reading the envelope can branch on
+  // it.
+  // The ai-config branch tracks its remediations on checks.ai_config
+  // (fix_applied / fix_failed are nested there, not on out.summary), so
+  // include that signal when deciding whether ANY fix actually ran.
+  // Without this, doctor --ai-config --fix applying chmod/icacls would
+  // simultaneously report checks.ai_config.fix_applied > 0 AND
+  // summary.fix_status: "already_present", which is contradictory.
+  const aiConfigFixed = !!(checks.ai_config && ((checks.ai_config.fix_applied || 0) > 0 || (checks.ai_config.fix_failed || 0) > 0));
+  if (args.fix && !out.summary.fix_applied && !out.summary.fix_attempted && !out.summary.fix_partial && !out.summary.fix_decline_reason && !aiConfigFixed) {
+    out.summary.fix_status = "already_present";
+    out.summary.fix_skipped_reason = "Signing key + skill signatures are already valid; nothing to remediate.";
+  }
+
   if (wantJson) {
     emit(out, indent);
     if (!allGreen) process.exitCode = EXIT_CODES.GENERIC_FAILURE;
@@ -6578,7 +6754,12 @@ function cmdDoctor(runner, args, runOpts, pretty) {
   mark(checks.cves, c => {
     if (!c.ok) return `CVE catalog FAILED (exit=${c.exit_code ?? "?"})`;
     const total = c.total ?? "?";
-    const breakdown = (c.cve_count != null && c.mal_count != null)
+    // Audit 3 B.2: enumerate every prefix so the sum equals total.
+    // Falls back to the legacy CVE + MAL display only if by_prefix isn't
+    // present (older catalog read paths).
+    const breakdown = c.by_prefix
+      ? ` (${Object.entries(c.by_prefix).sort().map(([p, n]) => `${n} ${p}`).join(" + ")})`
+      : (c.cve_count != null && c.mal_count != null)
       ? ` (${c.cve_count} CVE + ${c.mal_count} MAL)`
       : "";
     return `CVE catalog: ${total} entries${breakdown}, drift ${c.drift ?? 0}`;
@@ -6644,6 +6825,16 @@ function cmdDoctor(runner, args, runOpts, pretty) {
       ? ` (${c.policy_skips.length} judgement-shaped playbooks intentionally without a collector — see AGENTS.md)`
       : "";
     lines.push(`  ${icon} collector layer: ${c.with_collector ?? "?"}/${c.total_playbooks ?? "?"} playbooks have collectors${skipNote}`);
+    // Enumerate the policy-skipped playbooks in text mode so it carries
+    // the same operator-actionable information as the JSON envelope.
+    // Pre-fix the count was visible but the names appeared only in the
+    // structured output, forcing operators to parse JSON to learn which
+    // playbooks are policy-skipped.
+    if (Array.isArray(c.policy_skips) && c.policy_skips.length > 0) {
+      const shown = c.policy_skips.slice(0, 5).join(", ");
+      const more = c.policy_skips.length > 5 ? `, … +${c.policy_skips.length - 5} more` : "";
+      lines.push(`       policy-skipped: ${shown}${more}`);
+    }
     if (Array.isArray(c.load_errors) && c.load_errors.length > 0) {
       lines.push(`       ${c.load_errors.length} collector(s) failed to load:`);
       for (const e of c.load_errors.slice(0, 5)) {
@@ -6660,7 +6851,10 @@ function cmdDoctor(runner, args, runOpts, pretty) {
     const fileCount = c.scanned_files ?? 0;
     lines.push(`  ${icon} AI-assistant config audit: scanned ${fileCount} file(s) across ${dirCount} dir(s) of ${(c.directories_inspected || []).length} candidate root(s); ${findings.length} finding(s)`);
     if (c.platform === "win32" && findings.length === 0 && fileCount > 0) {
-      lines.push(`       (on Windows, POSIX mode bits are not load-bearing — manual ACL review noted for any sensitive files found)`);
+      lines.push(`       (Windows: ACL inspected via icacls; every sensitive file restricted to the workstation user)`);
+    }
+    if (c.walk_truncated) {
+      lines.push(`       (walk truncated at ${c.walk_caps?.max_files || "?"} file(s) / depth ${c.walk_caps?.max_depth || "?"}; rerun under a narrower path if you need exhaustive coverage)`);
     }
     for (const f of findings.slice(0, 5)) {
       const sev = f.severity === "error" ? "[!!]" : f.severity === "warn" ? "[warn]" : "[info]";
@@ -7243,10 +7437,10 @@ function cmdAsk(runner, args, runOpts, pretty) {
   // Keeps cmdAsk dependency-free; rich enough to cover the 80% of natural
   // queries listed in the operator report.
   const SYNONYMS = {
-    "credential": ["secret", "key", "token", "password", "cred"],
-    "credentials": ["secret", "key", "token", "password", "cred"],
-    "api key": ["secret", "credential"],
-    "api keys": ["secret", "credential"],
+    "credential": ["secret", "key", "token", "password", "cred", "secrets"],
+    "credentials": ["secret", "key", "token", "password", "cred", "secrets"],
+    "api key": ["secret", "credential", "secrets"],
+    "api keys": ["secret", "credential", "secrets"],
     "supply chain": ["sbom", "dependency", "vendor", "package", "library", "publish"],
     "supply-chain": ["sbom", "dependency", "vendor", "package", "library", "publish"],
     "npm package": ["sbom", "dependency", "library", "publish"],
@@ -7262,8 +7456,48 @@ function cmdAsk(runner, args, runOpts, pretty) {
     "secret": ["credential", "key", "token", "env", "leak"],
     "secrets": ["credential", "key", "token", "env", "leak", "repo"],
     "config": ["configuration", "settings"],
+    // Audit 3 C.2: missing identity / phishing / SSO / BEC vocabulary.
+    "phish": ["identity-sso-compromise", "idp-incident", "sso", "bec"],
+    "phishing": ["identity-sso-compromise", "idp-incident", "sso", "bec"],
+    "phished": ["identity-sso-compromise", "idp-incident", "sso", "bec"],
+    "sso": ["identity-sso-compromise", "idp-incident", "okta", "azure-ad", "entra"],
+    "oauth": ["identity-sso-compromise", "idp-incident", "openid", "oidc"],
+    "saml": ["identity-sso-compromise", "idp-incident"],
+    "okta": ["identity-sso-compromise", "idp-incident", "sso"],
+    "entra": ["identity-sso-compromise", "idp-incident", "azure-ad", "sso"],
+    "bec": ["identity-sso-compromise", "phish"],
+    "deepfake": ["identity-sso-compromise", "phish", "ai-c2"],
+    // Famous-attack vocabulary maps to library-author / sbom / supply-chain.
+    "left-pad": ["library-author", "sbom", "supply-chain-recovery", "npm"],
+    "left pad": ["library-author", "sbom", "supply-chain-recovery", "npm"],
+    "event-stream": ["library-author", "sbom", "supply-chain-recovery", "npm"],
+    "shai-hulud": ["library-author", "sbom", "supply-chain-recovery", "npm"],
+    "ransomware": ["ransomware", "kernel", "runtime"],
+    "rogue": ["ai-c2", "llm-tool-use-exfil"],
+    "agentic": ["ai-c2", "llm-tool-use-exfil", "mcp"],
+    // Credential theft from a developer laptop is cred-stores territory.
+    "credential theft": ["cred-stores", "secrets"],
+    "cred theft": ["cred-stores", "secrets"],
+    "credential exfil": ["cred-stores", "llm-tool-use-exfil"],
+    "developer laptop": ["cred-stores", "hardening"],
   };
 
+  // Audit 3 C.1: stopwords filtered after synonym expansion so common
+  // English words don't drive false positives via raw substring matching.
+  // "the the the the" used to route to ai-api because "the" substring-hit
+  // "anthropic" in the haystack.
+  const STOPWORDS = new Set([
+    "the", "and", "for", "are", "but", "not", "you", "all", "can", "had",
+    "her", "was", "one", "our", "out", "day", "get", "has", "him", "his",
+    "how", "man", "new", "now", "old", "see", "two", "way", "who", "boy",
+    "did", "its", "let", "put", "say", "she", "too", "use", "any", "got",
+    "from", "this", "that", "with", "have", "they", "what", "your", "when",
+    "which", "would", "could", "should", "there", "their", "about", "into",
+    "than", "then", "them", "some", "more", "most", "very", "much", "such",
+    "been", "were", "want", "well", "back", "good", "make", "made", "take",
+    "took", "give", "gave", "find", "found", "know", "knew", "told", "ago",
+  ]);
+
   // Tokenize question (length >= 2, lowercase) + expand via synonyms.
   const baseTokens = q.split(/\W+/).filter(t => t.length >= 2);
   const expanded = new Set(baseTokens);
@@ -7275,13 +7509,16 @@ function cmdAsk(runner, args, runOpts, pretty) {
   for (const t of baseTokens) {
     if (SYNONYMS[t]) for (const s of SYNONYMS[t]) expanded.add(s);
   }
+  // Filter stopwords AFTER synonym expansion (synonym lookup may have
+  // pulled in canonical tokens via a stopword-adjacent multi-word phrase).
+  for (const sw of STOPWORDS) expanded.delete(sw);
   const tokens = [...expanded];
 
   const scored = [];
   for (const id of ids) {
     let pb;
     try { pb = runner.loadPlaybook(id); } catch { continue; }
-    const haystack = [
+    const haystackText = [
       pb._meta?.id || id,
       pb.domain?.name || "",
       pb.domain?.attack_class || "",
@@ -7297,8 +7534,14 @@ function cmdAsk(runner, args, runOpts, pretty) {
       pb.phases?.look?.collection_scope?.asset_scope || "",
       pb.phases?.look?.collection_scope?.time_window || "",
     ].join(" ").toLowerCase();
+    // Audit 3 C.1: tokenize the haystack and use Set membership instead
+    // of raw substring matching. The substring approach matched "the" in
+    // "authentication" and "got" inside larger words, allowing stopword
+    // and partial-word false positives. Tokenized membership matches
+    // whole tokens only.
+    const haystackTokens = new Set(haystackText.split(/\W+/).filter(t => t.length >= 2));
     let score = 0;
-    for (const t of tokens) if (haystack.includes(t)) score++;
+    for (const t of tokens) if (haystackTokens.has(t)) score++;
     // ID match counts double — "secrets" should map to the secrets playbook.
     if (tokens.some(t => (pb._meta?.id || id) === t)) score += 3;
     scored.push({ id: pb._meta?.id || id, score });

package/data/_indexes/_meta.json

@@ -1,10 +1,10 @@
 {
   "schema_version": "1.1.0",
-  "generated_at": "2026-05-22T07:28:48.846Z",
+  "generated_at": "2026-05-22T23:28:32.508Z",
   "generator": "scripts/build-indexes.js",
   "source_count": 54,
   "source_hashes": {
-    "manifest.json": "f02056365a264901e95631245efa002968acee2e6432ab2e5131b6b53eadcd81",
+    "manifest.json": "020990fd6a13ea3a2a0273e80d0607c1cf0fd8887f03bb48294d2568a22d2789",
     "data/atlas-ttps.json": "d296c1d3e71807c9279b731f047e57796e85137f186586743a8cdad214b408f9",
     "data/attack-techniques.json": "49b6010b317edd219def135171ea8f3b1bbf1e00e9c5a08bf7237215ff54e2c3",
     "data/cve-catalog.json": "a09c83af3f9679a7ea73935726a1ff9de2cab94b4ab6321fc017fc147747d7c3",

package/lib/refresh-external.js

@@ -1426,6 +1426,22 @@ async function main() {
   // against each other for tokens. The two modes produce the same report
   // structure; only wall-clock differs.
   const runOne = async (src) => {
+    // Audit 3 A.1: --air-gap was honored by GHSA/OSV at the source-module
+    // level, but kev/epss/nvd/rfc/pins fell through to their live-network
+    // branches when neither fixtures nor cacheDir was wired up. Refuse
+    // here so the air-gap guarantee holds uniformly across every source.
+    if (ctx.airGap && !ctx.fixtures?.[src.name] && !ctx.cacheDir) {
+      return {
+        src,
+        diff: {
+          status: "unreachable",
+          diffs: [],
+          errors: 0,
+          summary: `air-gap mode: ${src.name} skipped (no fixture or cache configured; would have made a live network call)`,
+          air_gap_blocked: true,
+        },
+      };
+    }
     let diff;
     try {
       diff = await src.fetchDiff(ctx);
@@ -1457,6 +1473,10 @@ async function main() {
       diffs: diff.diffs,
       applies_to: src.applies_to,
       report_only: !!src.report_only,
+      // Audit 3 A.1: thread the central-dispatch air-gap short-circuit
+      // marker through to the persisted report so stdout-parsing consumers
+      // and the regression test can verify the network refusal happened.
+      ...(diff.air_gap_blocked ? { air_gap_blocked: true } : {}),
     };
     if (opts.apply && diff.diffs.length > 0 && !src.report_only) {
       const r = await src.applyDiff(ctx, diff.diffs);

package/manifest.json

@@ -1,6 +1,6 @@
 {
   "name": "exceptd-security",
-  "version": "0.13.57",
+  "version": "0.13.59",
   "description": "AI security skills grounded in mid-2026 threat reality, not stale framework documentation",
   "homepage": "https://exceptd.com",
   "license": "Apache-2.0",
@@ -53,7 +53,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "lXhZgoIrrVloO3XaTvo/43AxZn4mwErstd7DR0O/oVhD3AOGODM4HqrageYEou9WKOdMEGP5mJNTjJsXdP5NDA==",
-      "signed_at": "2026-05-22T07:13:12.339Z",
+      "signed_at": "2026-05-22T23:28:31.011Z",
       "cwe_refs": [
         "CWE-125",
         "CWE-362",
@@ -117,7 +117,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "vSVqu4wBm+d68ujZmM6Rto/HzViCkE0gPUcv/MYE/bjFiqamf/s0On4kTOo1KIveV9cOwYNxiItaGEWlVkRFDg==",
-      "signed_at": "2026-05-22T07:13:12.341Z",
+      "signed_at": "2026-05-22T23:28:31.013Z",
       "cwe_refs": [
         "CWE-1039",
         "CWE-1426",
@@ -180,7 +180,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "RIgXKvolQjgJdnlrDnVOd90IOY1B7VHHZD/YJQRzouL+wUeOLclPrdK/EgEuFyiu7lR4bi+Pl6aGB9G9tOxYCQ==",
-      "signed_at": "2026-05-22T07:13:12.341Z",
+      "signed_at": "2026-05-22T23:28:31.013Z",
       "cwe_refs": [
         "CWE-22",
         "CWE-345",
@@ -226,7 +226,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-01",
       "signature": "RYOxeq/o3uTwTWq4H7RcdH2Aclg9UyCERfUH9Frwkzncsowg7LgxpaEDc3swTCv73HMEGbU8wVbXguZ4JxHUCQ==",
-      "signed_at": "2026-05-22T07:13:12.341Z"
+      "signed_at": "2026-05-22T23:28:31.014Z"
     },
     {
       "name": "compliance-theater",
@@ -257,7 +257,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "DneJCPKCPcoe6nQ82XptqSqNfSRdt1orKaO+o7K36YCciDrzwJb+1BuBLusPDtpcdDaGY0y0e+AqiTYJklhBAQ==",
-      "signed_at": "2026-05-22T07:13:12.342Z"
+      "signed_at": "2026-05-22T23:28:31.014Z"
     },
     {
       "name": "exploit-scoring",
@@ -286,7 +286,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "NA1hoQycvQhSUoG5rwlXX0mOVmGxoXRVezkELGEA2nZOdGis4gXkHT3O6Sfw7zxE4JuMrsCb65TEeOWk9WEPDg==",
-      "signed_at": "2026-05-22T07:13:12.342Z"
+      "signed_at": "2026-05-22T23:28:31.015Z"
     },
     {
       "name": "rag-pipeline-security",
@@ -323,7 +323,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "XgrzcA2brPhXrSTxrcLnJec0OpgGYJBoSTUlJ10UdePHffxqb9LTVGnfbmEk1ykQifXREZexui2bG7X/+eFfCQ==",
-      "signed_at": "2026-05-22T07:13:12.343Z",
+      "signed_at": "2026-05-22T23:28:31.015Z",
       "cwe_refs": [
         "CWE-1395",
         "CWE-1426"
@@ -380,7 +380,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "9+hZlZOqZdeACUmamQk66L5levZhhwnFXuYRhdT6Mce99eQaKT7wNfWq12hXQztkRcVRKaFH+a01zwJQwsRQCA==",
-      "signed_at": "2026-05-22T07:13:12.343Z",
+      "signed_at": "2026-05-22T23:28:31.016Z",
       "d3fend_refs": [
         "D3-CA",
         "D3-CSPP",
@@ -415,7 +415,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-01",
       "signature": "ciqhVloMWWXEigPZvvwoV2c54tEqsDqsoc+sS/mNTFFJk2H+tz2+XUrgfEPRuYw0FeyNB6/+27pL2NpKHzUqAg==",
-      "signed_at": "2026-05-22T07:13:12.343Z",
+      "signed_at": "2026-05-22T23:28:31.016Z",
       "cwe_refs": [
         "CWE-1188"
       ]
@@ -443,7 +443,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-01",
       "signature": "xiHAhhdufm9hCKU8PLiPE0MX65ej2F4OZwtlWLGLCiie9/km+Kiqbt192LcMvr94v83C98pb9wIaqFsFWft6AQ==",
-      "signed_at": "2026-05-22T07:13:12.344Z"
+      "signed_at": "2026-05-22T23:28:31.016Z"
     },
     {
       "name": "global-grc",
@@ -475,7 +475,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-01",
       "signature": "oYsSk35N2Uzq7MRofACykylcVwkgPhI4luWZ14vmQT+gUKLyZiKVOUJbe1+7lGl6BYPRN0sUDQ0f7S5Eu5w2Ag==",
-      "signed_at": "2026-05-22T07:13:12.344Z"
+      "signed_at": "2026-05-22T23:28:31.017Z"
     },
     {
       "name": "zeroday-gap-learn",
@@ -502,7 +502,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-01",
       "signature": "igRqYyU1unRFH40BsPyAR62SPrk8QZv8dPGb8S9O9EvLCNOZAzm3t+HdT/NKqzWHwrpomOzkkkyLfYI/0qTUDA==",
-      "signed_at": "2026-05-22T07:13:12.344Z"
+      "signed_at": "2026-05-22T23:28:31.017Z"
     },
     {
       "name": "pqc-first",
@@ -554,7 +554,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "vhc3wuQEro/86s1ro2b/KakUXg8QVnySYTBqA7ebzv9oeR2HYO5bvGEJp3oOHWtL37JDqcCAHYadSN/qxIyCCA==",
-      "signed_at": "2026-05-22T07:13:12.345Z",
+      "signed_at": "2026-05-22T23:28:31.018Z",
       "cwe_refs": [
         "CWE-327"
       ],
@@ -601,7 +601,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "MS35nWm8djfJGn4OOoT0JKJ2aO+Dkbb6wOOWJYvNZlRKT3UGA59o2gxg1JOnD20hb/RwxtkmCujhl2tuYSR+AQ==",
-      "signed_at": "2026-05-22T07:13:12.345Z"
+      "signed_at": "2026-05-22T23:28:31.018Z"
     },
     {
       "name": "security-maturity-tiers",
@@ -638,7 +638,7 @@
       ],
       "last_threat_review": "2026-05-01",
       "signature": "8Px1s2lDj10/Q6erwEQlXgUHM1+OTruUR8qAHPX7Oo3k/l69N6P9sm0PsafS9wDFtj9l5C/OiLiFgzMlMt6vBw==",
-      "signed_at": "2026-05-22T07:13:12.346Z",
+      "signed_at": "2026-05-22T23:28:31.018Z",
       "cwe_refs": [
         "CWE-1188"
       ]
@@ -673,7 +673,7 @@
       "framework_gaps": [],
       "last_threat_review": "2026-05-11",
       "signature": "WAu5fRirzSOcnnZsTx2d/JJZwa/LPpXCi+31qATTGLmoNuhyy81k3ooPe9kCM3E0CLMtvTePg9DagYqBninZDQ==",
-      "signed_at": "2026-05-22T07:13:12.346Z"
+      "signed_at": "2026-05-22T23:28:31.019Z"
     },
     {
       "name": "attack-surface-pentest",
@@ -744,7 +744,7 @@
         "PTES revision incorporating AI-surface enumeration"
       ],
       "signature": "7eEwCXFd9pDKUw7yCUbRJSjfzozE44dwwwemCQUPm8JBPztLltibD9bL/RszSbYyCrYJmVb5Drncz2cGe62gCw==",
-      "signed_at": "2026-05-22T07:13:12.346Z"
+      "signed_at": "2026-05-22T23:28:31.019Z"
     },
     {
       "name": "fuzz-testing-strategy",
@@ -804,7 +804,7 @@
         "OSS-Fuzz-Gen / AI-assisted harness generation becoming the default expectation for OSS maintainers"
       ],
       "signature": "Z7ypCUnXx8JpLtgxxB6RHNi39w74AmrGY1N4ofAGCXhkuM2EaFVm1AU0dvl9UQ1bVLfHKEDGqMO/TwlIY7RABg==",
-      "signed_at": "2026-05-22T07:13:12.347Z"
+      "signed_at": "2026-05-22T23:28:31.019Z"
     },
     {
       "name": "dlp-gap-analysis",
@@ -879,7 +879,7 @@
         "Quebec Law 25, India DPDPA, KSA PDPL enforcement actions naming AI-tool prompt data as in-scope personal information"
       ],
       "signature": "fgxG344JGYBWWWwFXZ1IzGipWKP7EyBhrsvsbsb0CCGXfv/MvNHVNI6G0zQddCsWX1JeQbhZT3Vk8v1uJKDTDA==",
-      "signed_at": "2026-05-22T07:13:12.347Z"
+      "signed_at": "2026-05-22T23:28:31.020Z"
     },
     {
       "name": "supply-chain-integrity",
@@ -956,7 +956,7 @@
         "OpenSSF model-signing — emerging Sigstore-based signing standard for ML model weights; track for production adoption"
       ],
       "signature": "pcLrM98A3vUSZRjwNAk0aZ9umvOwB41XCLLsCOy/IebB2F/06oIrGUKkMHtHwm4pTVPShMMcKdZQQ3jz30FnCg==",
-      "signed_at": "2026-05-22T07:13:12.347Z"
+      "signed_at": "2026-05-22T23:28:31.020Z"
     },
     {
       "name": "defensive-countermeasure-mapping",
@@ -1013,7 +1013,7 @@
       ],
       "last_threat_review": "2026-05-11",
       "signature": "gqF8eU3VBrZhO2WnlcqKa7wm1d2mmWtvpbmx0kNCgHojNV+qEt+Ij84RO6bZvaUqhfYPWizWL79Fa4DL0curAQ==",
-      "signed_at": "2026-05-22T07:13:12.347Z"
+      "signed_at": "2026-05-22T23:28:31.020Z"
     },
     {
       "name": "identity-assurance",
@@ -1080,7 +1080,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "Wv5hGMeHjlaQK1zwicVCA7AvdKgJBgvcjdpGM9Ywahh9tagAKhbkOjybowDQZzu7OZ3bDkbh6pBYc1Sdwr6NAA==",
-      "signed_at": "2026-05-22T07:13:12.348Z"
+      "signed_at": "2026-05-22T23:28:31.021Z"
     },
     {
       "name": "ot-ics-security",
@@ -1136,7 +1136,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "8t5qKHd3yWi57dvG36YQkLN/X9bQWqtEiYjay4IfSmqhJpM/xXPaQVKNGz3wscrO8OLKUZ0OaX7Mj5kzpgBKBQ==",
-      "signed_at": "2026-05-22T07:13:12.348Z"
+      "signed_at": "2026-05-22T23:28:31.021Z"
     },
     {
       "name": "coordinated-vuln-disclosure",
@@ -1188,7 +1188,7 @@
         "NYDFS 23 NYCRR 500 amendments potentially adding explicit CVD program requirements"
       ],
       "signature": "GDGt4UPqBa04PjlpSmpyihGzd3OgfBN7jaAK5tfwp+LRSs3ygKOdbeivUCCHNagTY1hE6hG2Ou40ADfBFuXeAg==",
-      "signed_at": "2026-05-22T07:13:12.348Z"
+      "signed_at": "2026-05-22T23:28:31.021Z"
     },
     {
       "name": "threat-modeling-methodology",
@@ -1238,7 +1238,7 @@
         "PASTA v2 updates incorporating AI/ML application threats"
       ],
       "signature": "rFBpOQEJUPpl+v88Lw/WqVJRhTl80vy0VbPAbzQj3Q0suJRRrJg368I9uKu5LXIBKFDvKxnGIcIzbGg9NUtaCA==",
-      "signed_at": "2026-05-22T07:13:12.349Z"
+      "signed_at": "2026-05-22T23:28:31.022Z"
     },
     {
       "name": "webapp-security",
@@ -1312,7 +1312,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "ux85YI4t2mVHOyt744Yin1HHy+z11JIFygjKfFfQOBBl5QVV3A267jeIy7utix85irMcpZm/T3yx/ooqiK2tBA==",
-      "signed_at": "2026-05-22T07:13:12.349Z"
+      "signed_at": "2026-05-22T23:28:31.022Z"
     },
     {
       "name": "ai-risk-management",
@@ -1362,7 +1362,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "IIXnkZ5ZNqFwOto5KfytADTLLZLoyXNZACD1ORZ40P1HUAQxe6u2uyXFzzsfuob4Uy06jNkRGr2FFgCphUH1Cw==",
-      "signed_at": "2026-05-22T07:13:12.349Z"
+      "signed_at": "2026-05-22T23:28:31.022Z"
     },
     {
       "name": "sector-healthcare",
@@ -1422,7 +1422,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "AhF9KF8ZBlDteciV+F8IBSmFVYCvQOn44GmD4rZjgLoPxfIv/QE1/vSkK32zyqDKtHWkLSXExbkkPkxA/V6dDw==",
-      "signed_at": "2026-05-22T07:13:12.350Z"
+      "signed_at": "2026-05-22T23:28:31.023Z"
     },
     {
       "name": "sector-financial",
@@ -1503,7 +1503,7 @@
         "TIBER-EU framework v2.0 alignment with DORA TLPT RTS (JC 2024/40); cross-recognition with CBEST and iCAST"
       ],
       "signature": "HQgZvb4ReziEz5rNFr8i/O8/rJEZR+iHRROT7m/D2QUqhrcNISPkYXENsUZlG8xapzy/Ik92ehkseyj4hdmhCQ==",
-      "signed_at": "2026-05-22T07:13:12.350Z"
+      "signed_at": "2026-05-22T23:28:31.023Z"
     },
     {
       "name": "sector-federal-government",
@@ -1572,7 +1572,7 @@
         "Australia PSPF 2024 revision and ISM quarterly updates — track for Essential Eight Maturity Level requirements for federal entities"
       ],
       "signature": "linxmsXZiOYtcs71sSWgGCrvb8xQfmxmtTY5PRvZJ0/8FgJulo0tQtejzexYG775s7XhjAmGsDP238BQTQ8ADA==",
-      "signed_at": "2026-05-22T07:13:12.351Z"
+      "signed_at": "2026-05-22T23:28:31.024Z"
     },
     {
       "name": "sector-energy",
@@ -1637,7 +1637,7 @@
         "ICS-CERT advisory feed (https://www.cisa.gov/news-events/cybersecurity-advisories/ics-advisories) for vendor CVEs in Siemens, Rockwell, Schneider Electric, ABB, GE Vernova, Hitachi Energy, AVEVA / OSIsoft PI"
       ],
       "signature": "JjBfc0ovta560Clk0x3QGRM5osFJDwcvpy3rT7QEGdCIL827jzE8QCow1C8deXq+4JhY2sA/d7/8IsxikdlkCg==",
-      "signed_at": "2026-05-22T07:13:12.352Z"
+      "signed_at": "2026-05-22T23:28:31.024Z"
     },
     {
       "name": "sector-telecom",
@@ -1723,7 +1723,7 @@
         "O-RAN SFG / WG11 security specifications"
       ],
       "signature": "JWVxKFoKrbX4d+Tko1d4OBdwyg25MfFFKn4CT6E/CzH+YwnU3T6Y76uBQIKg3+gIGTvPduqyvQwQQ5FxKDuPBw==",
-      "signed_at": "2026-05-22T07:13:12.352Z"
+      "signed_at": "2026-05-22T23:28:31.025Z"
     },
     {
       "name": "api-security",
@@ -1792,7 +1792,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "BmCRCestWqr55+fCynEhtAl5NWLT+xLTkpwS0Icp3SaoZOw/ce3Y6TtqjHRSKn4CBJq7YDiLRWxmhO3MStvOAA==",
-      "signed_at": "2026-05-22T07:13:12.352Z"
+      "signed_at": "2026-05-22T23:28:31.025Z"
     },
     {
       "name": "cloud-security",
@@ -1873,7 +1873,7 @@
         "CISA KEV additions for cloud-control-plane CVEs (IMDSv1 abuses, federation token mishandling, cross-tenant boundary failures); CISA Cybersecurity Advisories for cross-cloud advisories"
       ],
       "signature": "/DV3pmZwrRySrk1OCbyI+0BQESacjupJfUX3eC2NGtXuYOBro0vndIP+z27heFxumnjU3a9sfla7/U9X+pqnDw==",
-      "signed_at": "2026-05-22T07:13:12.352Z"
+      "signed_at": "2026-05-22T23:28:31.025Z"
     },
     {
       "name": "container-runtime-security",
@@ -1935,7 +1935,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "E2UGSf9ATyYgzBr8uM/0ubOUmDqo1jVA7f9mVxv6LHfWGCNuQNXDyuNou9VAmUCeeXEeUYIi3AFjXkJqpOkxDA==",
-      "signed_at": "2026-05-22T07:13:12.353Z"
+      "signed_at": "2026-05-22T23:28:31.026Z"
     },
     {
       "name": "mlops-security",
@@ -2006,7 +2006,7 @@
         "MITRE ATLAS v5.6.0 (released February 2026) shipped the AML.T0010 sub-technique expansion this forecast tracked plus new techniques (\"Publish Poisoned AI Agent Tool\", \"Escape to Host\"); inventory now 16 tactics, 84 techniques, 56 sub-techniques. Forward watch: ATLAS v5.5 / v6.0 — track next-cadence updates to agentic-AI TTPs and MLOps-pipeline-specific techniques"
       ],
       "signature": "BGNE6ZQWBA1LmsUFe8tU0L67iGDSrFqiuqaZD2f1KqfcyqqzQfMs9PWNHFzxxaJmXeKlm87eU8lgELF0bX+RBA==",
-      "signed_at": "2026-05-22T07:13:12.353Z"
+      "signed_at": "2026-05-22T23:28:31.026Z"
     },
     {
       "name": "incident-response-playbook",
@@ -2068,7 +2068,7 @@
         "NYDFS 23 NYCRR 500.17 amendments tightening ransom-payment 24h disclosure operationalization"
       ],
       "signature": "FkZQerh3VHVJAwIcCktDyMRh5KE2+Em/i0ek8zEz7JG/PXtQx8ujHWTh3VjZbOLhPNtdB2qxgXOIAYIofaVOAQ==",
-      "signed_at": "2026-05-22T07:13:12.354Z"
+      "signed_at": "2026-05-22T23:28:31.027Z"
     },
     {
       "name": "ransomware-response",
@@ -2148,7 +2148,7 @@
       ],
       "last_threat_review": "2026-05-15",
       "signature": "n3UToNuN3A1HgLvcuqmIx8vrZY71+r/79waK92jG+rSX4uYOzkmxMUpROrE5K9bDwMezNBHdjWv8Uul6zugyDQ==",
-      "signed_at": "2026-05-22T07:13:12.354Z"
+      "signed_at": "2026-05-22T23:28:31.027Z"
     },
     {
       "name": "email-security-anti-phishing",
@@ -2201,7 +2201,7 @@
       "d3fend_refs": [],
       "last_threat_review": "2026-05-11",
       "signature": "rK+WnuS+9tqEABmwc0jO/PEmxcLjG1/tmUb897HsClQeKzf+TQOlwBE+OsbtuKxpjYNwur62Xxs3TxObkwm8Cw==",
-      "signed_at": "2026-05-22T07:13:12.354Z"
+      "signed_at": "2026-05-22T23:28:31.027Z"
     },
     {
       "name": "age-gates-child-safety",
@@ -2269,7 +2269,7 @@
         "US state adult-site age-verification laws — 19+ states by mid-2026 (TX HB 18 upheld by SCOTUS June 2025 in Free Speech Coalition v. Paxton); track ongoing challenges in remaining states"
       ],
       "signature": "+OO0RhQ303RJV7kaH38IuZpLeQbapep6Ds4Re/WEZu0FHBwKSlwvF7jbtP7KQ57xldJYn/xZm2jaszyOacMfDg==",
-      "signed_at": "2026-05-22T07:13:12.355Z"
+      "signed_at": "2026-05-22T23:28:31.028Z"
     },
     {
       "name": "cloud-iam-incident",
@@ -2349,7 +2349,7 @@
       ],
       "last_threat_review": "2026-05-15",
       "signature": "e/kij7GtKaytROyIj7V5RH+FC9WtmVFzrmG2kIlNDNn29ep/CRNlIQKwXLpzo/81AIf634pmdr1qy/+vwIuUDA==",
-      "signed_at": "2026-05-22T07:13:12.355Z"
+      "signed_at": "2026-05-22T23:28:31.028Z"
     },
     {
       "name": "idp-incident-response",
@@ -2430,11 +2430,11 @@
       ],
       "last_threat_review": "2026-05-15",
       "signature": "ew9Kglc9fAZzbn0ZIfGP7WSK/j4eV2VhSvpy+s5bEfNEVYIMa2kZjnGBapgUsyGDLes9H9K2ovjQyX17+GKiBw==",
-      "signed_at": "2026-05-22T07:13:12.355Z"
+      "signed_at": "2026-05-22T23:28:31.028Z"
     }
   ],
   "manifest_signature": {
     "algorithm": "Ed25519",
-    "signature_base64": "Mqdh4G/iwFRFGu8rZzRnFr9lRtUZuHwTLJhvCMik7tfut7rHh2YKeGKojYcQ//dM7KISIHkkq3wnC+zJzWcpCw=="
+    "signature_base64": "jYdDFzIsg9EdwlNtBAT5NYGZYTtQzpZxo5lGAUJ4lG2hSWgmFy8dOqe4fL1o++uqDYDbUfh8tF/CkDOiGqJrCg=="
   }
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blamejs/exceptd-skills",
-  "version": "0.13.57",
+  "version": "0.13.59",
   "description": "AI security skills grounded in mid-2026 threat reality, not stale framework documentation. 42 skills, 10 catalogs (312 CVEs / 171 CWEs / 805 ATT&CK + ICS / 170 ATLAS / 468 D3FEND / 7476 RFCs), 34 jurisdictions, 10-class catalog gap detector + budget gate, real XML parser + canonical-form diff + content-pattern regression detection, Ed25519-signed.",
   "keywords": [
     "ai-security",

package/sbom.cdx.json

@@ -1,22 +1,22 @@
 {
   "bomFormat": "CycloneDX",
   "specVersion": "1.6",
-  "serialNumber": "urn:uuid:b0a6b327-121e-40c6-8092-871c7183b709",
+  "serialNumber": "urn:uuid:eb996267-363d-46e0-9a5c-f259a8c46d9b",
   "version": 1,
   "metadata": {
-    "timestamp": "2119-12-02T06:07:35.000Z",
+    "timestamp": "2151-04-04T19:39:19.000Z",
     "tools": [
       {
         "vendor": "blamejs",
         "name": "scripts/refresh-sbom.js",
-        "version": "0.13.57"
+        "version": "0.13.59"
       }
     ],
     "component": {
-      "bom-ref": "pkg:npm/@blamejs/exceptd-skills@0.13.57",
+      "bom-ref": "pkg:npm/@blamejs/exceptd-skills@0.13.59",
       "type": "application",
       "name": "@blamejs/exceptd-skills",
-      "version": "0.13.57",
+      "version": "0.13.59",
       "description": "AI security skills grounded in mid-2026 threat reality, not stale framework documentation. 42 skills, 10 catalogs (312 CVEs / 171 CWEs / 805 ATT&CK + ICS / 170 ATLAS / 468 D3FEND / 7476 RFCs), 34 jurisdictions, 10-class catalog gap detector + budget gate, real XML parser + canonical-form diff + content-pattern regression detection, Ed25519-signed.",
       "licenses": [
         {
@@ -25,17 +25,17 @@
           }
         }
       ],
-      "purl": "pkg:npm/%40blamejs/exceptd-skills@0.13.57",
+      "purl": "pkg:npm/%40blamejs/exceptd-skills@0.13.59",
       "hashes": [
         {
           "alg": "SHA-256",
-          "content": "699e406210e768c36fb8c34f9c860c744a6e34e39eb3e2faec853b870a65a0b4"
+          "content": "97387465e487f40271bdfd6dc20958446c2772c571c9c7671e9f912e0a6b32ae"
         }
       ],
       "externalReferences": [
         {
           "type": "distribution",
-          "url": "https://www.npmjs.com/package/@blamejs/exceptd-skills/v/0.13.57"
+          "url": "https://www.npmjs.com/package/@blamejs/exceptd-skills/v/0.13.59"
         },
         {
           "type": "vcs",
@@ -116,11 +116,11 @@
       "hashes": [
         {
           "alg": "SHA-256",
-          "content": "95bd558e91d2239599d30e889edd3313493776362fb57cb4b76b153b6ee2792a"
+          "content": "d08eca001f3bd9547d899b0531b59b5bcf929bcc1134f5f920bc85453a96bee8"
         },
         {
           "alg": "SHA3-512",
-          "content": "585290d78871e4effc0465ddbfdf5316466c6a1719d2df68faa8521d0445b759269ec27f12726943ef10eb5ce463702d1d883aa53ca82743dda335c416af4ec9"
+          "content": "9b6e0d750008f8b273c44016fd002d31f38ebb95b00787cd57081c29f8afed8c9c56b94a85c17b9a1bd4edacdbf12ba602acda6459d7743ab409ff3471114932"
         }
       ]
     },
@@ -281,11 +281,11 @@
       "hashes": [
         {
           "alg": "SHA-256",
-          "content": "75b10d80c2b634e132427542056cd2ca67dba954141e6d767677a1ab18ffac9d"
+          "content": "679a543aec746f1e99d891099bd04ab88fd0b8ba6e20a3484c9d9e3a9d33fd7e"
         },
         {
           "alg": "SHA3-512",
-          "content": "70e8c7016d5be30a646de75e7f2aaa7b9999277ecbfbaea660ff546ece14e4dfe513af956ccacbe56f67ae3cf052148e863526a72ac93ccd0e3d76b888aa503b"
+          "content": "694240b8a9268abc9ec54bdb52df56389fb115a81f99d70121ccb7e427d7bc670d7d4651fbaaec9b7be35f03c861b75b06177ba4e372890304542aa1a4615a83"
         }
       ]
     },
@@ -1271,11 +1271,11 @@
       "hashes": [
         {
           "alg": "SHA-256",
-          "content": "edf7fe483cffe4fe0429ff6f43cd1fec1b35a66d39d4bbd2ae7298886c3af3b9"
+          "content": "f769111b2412cd0ae3ff6f2adc54002355a271262d79a2c645a4e0f41e2a1c1a"
         },
         {
           "alg": "SHA3-512",
-          "content": "3bffc99ca6c2120b67e004fb100f9fb8674831a1d201fc4b93061933b69b31bef6cd3a90b4dc77fb554c023b88cfda8969a4a4a53a0ec6faa5171dfb8f4b044b"
+          "content": "5f58125b5706702ad3c6f1c4738772d8a61f6183426af178ae3666f16c8201c623920f668f0b83075484bfd1b100b83563161b7e36e6ffba7c073af5dadbb4b9"
         }
       ]
     },
@@ -1661,11 +1661,11 @@
       "hashes": [
         {
           "alg": "SHA-256",
-          "content": "f02056365a264901e95631245efa002968acee2e6432ab2e5131b6b53eadcd81"
+          "content": "020990fd6a13ea3a2a0273e80d0607c1cf0fd8887f03bb48294d2568a22d2789"
         },
         {
           "alg": "SHA3-512",
-          "content": "3f2f83b42aceb51653aa005f11404db852ca9d043536eb626cd03240f15ef92ec5a353fea1024518e149a59ac2e2b38623417c7b49907240d484686975d1e95f"
+          "content": "4bbf18c5ec14dbae987465ebca8dc06b2fcce0347eae3f834fb1fc33115289acf6484385b25f728548b70e37a7f4f2ce3d4c314b36465961cc22dbaf7c974e14"
         }
       ]
     },