@blamejs/exceptd-skills 0.18.6 → 0.18.7

package/CHANGELOG.md

@@ -1,5 +1,25 @@
 # Changelog
 
+## 0.18.7 — 2026-06-20
+
+Escalation and feeds_into conditions that use an `any`/`all` quantifier prefix, a quoted `matches '…'` pattern, or a quoted `contains` member now evaluate instead of silently failing closed. Several catalog chains that depended on these forms — most visibly the compliance-theater → SBOM correlation — were dead and now fire; the feeds_into evaluation context also exposes the bare `blast_radius_score`, `theater_verdict`, and `compliance_theater_check` tokens the catalog references, so an engine-computed value is no longer lost to a `null` lookup. The SBOM → kernel / MCP / AI-API follow-ups now fire too: a matched CVE carries an `attack_class`, the analyze and close contexts expose the matched-CVE array the `any matched_cve.attack_class == …` rules quantify over, and the CVEs that map unambiguously to one of those deep-dive classes are tagged in the catalog — so an SBOM scan that surfaces an MCP-server RCE or a prompt-injection-to-RCE routes the operator into the matching deep dive, while an unclassified match routes nowhere rather than guessing. `IN […]` member lists, AND/OR joins, and outer parentheses are parsed quote-aware, so a comma, bracket, or operator inside a quoted value no longer tears the clause, and a `contains`/`IN` against a path that doesn't resolve records a diagnostic instead of evaluating to an invisible false.
+
+RWEP scoring no longer silently drops the active-exploitation weight when a CVE's `active_exploitation` value is outside the recognised vocabulary or merely stray-cased (`Confirmed`, ` CONFIRMED `): canonical values are case-normalised, and a genuinely unrecognised value contributes zero **and** emits a warning so the dropped weight is visible rather than quietly lowering the priority. Curation now rejects a mixed-shape `rwep_factors` block (post-weight integers with a stray non-numeric value) at apply time instead of letting it route through the wrong scoring path and write a silently-wrong RWEP to the catalog. The scorer's two factor shapes no longer disagree on that block — the derive path applies the same mixed-shape guard the validator does, so it can't fall back to the boolean path and under-score — the Shape-B sum clamps each weight to its ceiling, and an SLA timeline is derived only from a real number rather than treating a NaN as the least-urgent tier.
+
+`attest diff` no longer manufactures phantom drift when one attestation carries real evidence and the other is empty — the full-catalog placeholder now stands in only when both sides are empty, so a populated side diffs against nothing and yields the genuine added/removed instead of every catalog id reading as drift. It also no longer masks a real change when an artifact's secret or path is carried under a key other than `value`: that evidence is preserved through capture and compared, so two attestations recording different secrets at different paths no longer diff as unchanged.
+
+An empty-string flag value is rejected rather than silently degrading the command. `--evidence ""`, `--evidence-dir ""`, `--phase ""`, `--against ""`, `--playbook ""`, and `--session-id ""` now error instead of running with no scope or bypassing their phase / target / dispatcher validation, and `attest export` redacts `signal_overrides` in its labeled output instead of emitting them verbatim.
+
+The secrets collector no longer double-reports an Anthropic `sk-ant-…` key as both an OpenAI and an Anthropic key, and now surfaces subtrees skipped for exceeding the depth cap instead of leaving them silently unscanned. The collectors and the CVE-cache reader read each file through to end-of-file, so a key or misconfiguration sitting past a short-read boundary on a network or sync-backed filesystem is no longer dropped and the scan no longer reports clean over content it never decoded. The library-author collector no longer flags npm-workspaces lockfile links (`link: true`) as missing integrity — a false positive on every workspaces monorepo. The CI/CD collector matches the GitHub OIDC issuer and a `GITHUB_TOKEN` secret name exactly, so a look-alike issuer host or a custom `GITHUB_TOKEN_*` secret is no longer mis-handled.
+
+The network-backed pollers (OSV, GHSA, the Datatracker freshness probe, the package prefetch, and the watchlist org scan) retry transient failures with backoff and cap response size instead of buffering an unbounded body into memory; the prefetch retry classifier reads the error status under the field the retry helper sets, so a 429/5xx is retried rather than treated as fatal. The vendored-tree integrity check fails closed when a license hash or a per-file hash is absent — a stripped hash no longer reads as verified.
+
+CVE-catalog validation rejects impossible calendar dates (for example `2026-02-30`, or Feb 29 in a non-leap year) on KEV deadline fields instead of letting them roll over to a valid-looking date, and the public-exploit URL classifier is anchored so a look-alike host such as `exploit-db.com.attacker.example` is no longer treated as a public-exploit source.
+
+The index-freshness gate now verifies that every derived index file still exists and parses, not only that the source hashes match — a deleted or truncated index no longer passes as current. `refresh --network` and the vendored-file online validator pin their fetches to the npm registry / GitHub hosts, so a tampered metadata response or redirect cannot steer a download at an arbitrary address. `refresh --network` additionally verifies the npm registry's signature over the fetched tarball and authenticates the `data/` tree and manifest snapshot before swapping them in, so a forged metadata response that recomputes the transport hashes is still refused. The live CVE-validation downgrade guard now reads the upstream metric's declared CVSS version, so a legacy v2 score NVD returns without a vector string is still recognised as a downgrade and held back from a curated v3.x entry — matching the cached refresh path, which already used the declared version.
+
+Internal: the diff-coverage export extractor and the codebase-pattern detectors are now string-aware, so a brace or `//` inside a string literal can no longer hide a new export from the coverage gate or disarm the process-exit / dynamic-regex detectors. Temp files in the test suite are created inside owner-only directories. The signing-key bootstrap creates each key with an exclusive (`O_EXCL`) open so an existing key is refused atomically rather than via a separate existence check, and the index validator opens derived files with `O_NOFOLLOW` and reads them through the descriptor, removing the check-then-read windows static analysis flagged. The signer schema-validates the manifest before writing a signature, so it can no longer bless a manifest the verifier would reject at install time. Several gate detectors close blind spots: the test-count gate ignores `test(` inside a multiline template literal, the export extractor sees method-shorthand and single-identifier `module.exports`, the process-exit detector covers `scripts/` and `bin/`, the version-tag detector no longer counts an IPv4 octet run, and `build-indexes --changed` regenerates a deleted index instead of trusting a clean source hash and no longer rewrites a timestamp on a no-op run.
+
 ## 0.18.6 — 2026-06-14
 
 `attest prune` now ages out sessions that hold only replay records (whose attestation was removed). Such a session had no datable timestamp, so it was kept indefinitely and the attestation store could grow without bound; it is now dated by its newest replay timestamp and pruned past the cutoff like any other session.

package/bin/exceptd.js

@@ -60,7 +60,7 @@ const PKG_ROOT = path.resolve(__dirname, "..");
 // constants so a new verb cannot regress the exit-code contract by typo,
 // and so the help-text dump (`doctor --exit-codes`) and the runtime
 // behavior share the same source of truth.
-const { EXIT_CODES, listExitCodes } = require(path.join(PKG_ROOT, "lib", "exit-codes.js"));
+const { EXIT_CODES, listExitCodes, safeExit } = require(path.join(PKG_ROOT, "lib", "exit-codes.js"));
 const { validateIdComponent } = require(path.join(PKG_ROOT, "lib", "id-validation.js"));
 const { suggestFlag, flagsFor, VERB_FLAG_ALLOWLIST } = require(path.join(PKG_ROOT, "lib", "flag-suggest.js"));
 const codepointClass = require(path.join(PKG_ROOT, "vendor", "blamejs", "codepoint-class.js"));
@@ -612,7 +612,7 @@ function main() {
   }
   if (cmd === "version" || cmd === "--version" || cmd === "-v") {
     process.stdout.write(readPkgVersion() + "\n");
-    process.exit(0);
+    safeExit(EXIT_CODES.SUCCESS); return;
   }
   if (cmd === "path") {
     // v0.11.14 (#130): `path copy` was silently consuming the `copy` arg and
@@ -638,14 +638,14 @@ function main() {
       if (copied) {
         process.stderr.write(`[exceptd path] copied to clipboard: ${PKG_ROOT}\n`);
         process.stdout.write(PKG_ROOT + "\n");
-        process.exit(0);
+        safeExit(EXIT_CODES.SUCCESS); return;
       }
       process.stderr.write(`[exceptd path] copy: no clipboard tool available (tried: ${tried.join(", ")}). Path printed to stdout instead.\n`);
       process.stdout.write(PKG_ROOT + "\n");
-      process.exit(0);
+      safeExit(EXIT_CODES.SUCCESS); return;
     }
     process.stdout.write(PKG_ROOT + "\n");
-    process.exit(0);
+    safeExit(EXIT_CODES.SUCCESS); return;
   }
 
   // v0.13.0: hard-refuse the v0.10.x legacy verbs that were
@@ -1157,9 +1157,9 @@ function hasReadableStdin() {
  * on failure so the caller can wrap it with its own verb prefix.
  */
 const ISO_DATE_RE = /^\d{4}-\d{2}-\d{2}(?:[T ]\d{2}:\d{2}(?::\d{2}(?:\.\d+)?)?(?:Z|[+-]\d{2}:?\d{2})?)?$/;
-function validateIsoSince(raw) {
+function validateIsoSince(raw, flagName = "--since") {
   if (typeof raw !== "string" || !ISO_DATE_RE.test(raw) || isNaN(Date.parse(raw))) {
-    return `--since must be a parseable ISO-8601 calendar timestamp (e.g. 2026-05-01 or 2026-05-01T00:00:00Z). Got: ${JSON.stringify(String(raw)).slice(0, 80)}`;
+    return `${flagName} must be a parseable ISO-8601 calendar timestamp (e.g. 2026-05-01 or 2026-05-01T00:00:00Z). Got: ${JSON.stringify(String(raw)).slice(0, 80)}`;
   }
   return null;
 }
@@ -1207,7 +1207,7 @@ function detectVexShape(doc) {
   // OpenVEX: @context starts with https://openvex.dev AND statements[]
   const ctx = doc["@context"];
   const ctxStr = Array.isArray(ctx) ? ctx[0] : ctx;
-  if (typeof ctxStr === "string" && ctxStr.startsWith("https://openvex.dev") && Array.isArray(doc.statements)) {
+  if (typeof ctxStr === "string" && ctxStr.startsWith("https://openvex.dev/") && Array.isArray(doc.statements)) {
     return { ok: true, detected: "openvex", top_level_keys: keys };
   }
   // Common false-positive shapes — give the operator a hint.
@@ -1399,13 +1399,20 @@ function dispatchPlaybook(cmd, argv) {
     );
     process.env.EXCEPTD_AIR_GAP_NOTICE_SHOWN = "1";
   }
-  if (args["session-id"]) {
+  if (args["session-id"] !== undefined) {
     // --session-id is a filesystem path component (resolves to
     // .exceptd/attestations/<id>/attestation.json). Operator-supplied input
     // with `..` or path separators escapes the attestation root. Route
     // through the shared validateIdComponent('session') helper so the regex
     // + all-dots refusal stay aligned with persistAttestation /
     // validateSessionIdForRead.
+    //
+    // Presence-gated (`!== undefined`), not truthy-gated: `--session-id ""`
+    // / `--session-id=` carry an explicit empty value the operator meant to
+    // pin. A truthy gate skips the validator for "", silently substituting a
+    // random id and discarding the operator's intent. validateIdComponent
+    // rejects "" with "must not be empty", matching the --operator empty
+    // refusal below.
     const sid = args["session-id"];
     const r = validateIdComponent(sid, "session");
     if (!r.ok) {
@@ -2497,6 +2504,11 @@ async function cmdCollect(runner, args, runOpts, pretty) {
   // --cwd <path> overrides process.cwd(). Validated as an existing
   // directory; non-existent / non-directory cwd is operator error.
   let cwd = process.cwd();
+  // An explicit empty value (`--cwd ""`) would otherwise be falsy and silently
+  // scan process.cwd() — the wrong directory — reported as a successful run.
+  if (args.cwd === "") {
+    return emitError(`collect: --cwd was given an empty value; pass an existing directory path`, { verb: "collect", playbook_id: playbookId }, pretty);
+  }
   if (args.cwd) {
     const resolved = path.resolve(String(args.cwd));
     let stat;
@@ -3358,6 +3370,12 @@ function cmdRun(runner, args, runOpts, pretty) {
   // first, then falls back to a strict isTTY===false check only on Windows
   // (where fstat on a pipe is unreliable). MSYS-bash on win32 reports
   // isTTY === false for genuine piped input, so that path still works.
+  // An explicit empty value (`--evidence ""`) is operator error: it would
+  // otherwise be falsy and silently produce a no-evidence "not_detected" run at
+  // exit 0, masking the fact that the intended evidence never loaded.
+  if (args.evidence === "") {
+    return emitError("run: --evidence was given an empty value; pass a file path, '-' for stdin, or omit --evidence for a no-evidence run", { verb: "run" }, pretty);
+  }
   const autoStdin = !args.evidence && hasReadableStdin();
   if (autoStdin) {
     args.evidence = "-";
@@ -4168,6 +4186,20 @@ function cmdRunMulti(runner, ids, args, runOpts, pretty, meta) {
   runOpts.session_id = sessionId;
 
   let bundle = {};
+  // An explicit empty value (--evidence "" / --evidence= / an unset shell
+  // variable) is falsy, so the truthiness-gated reads below would skip
+  // entirely, the bundle would stay {}, every playbook would run with no
+  // evidence, and the contract would report a clean not_detected at exit 0 —
+  // a false-clean that hides the fact the operator's intended evidence never
+  // loaded. Mirror the single-playbook run / ci empty-value guards: refuse
+  // the empty value loudly rather than running a vacuous contract. Presence,
+  // not truthiness, is the test.
+  if (args.evidence === "") {
+    return emitError("run: --evidence was given an empty value; pass a file path, '-' for stdin, or omit --evidence for a no-evidence run", { verb: "run", flag: "evidence" }, pretty);
+  }
+  if (args["evidence-dir"] === "") {
+    return emitError("run: --evidence-dir was given an empty value; pass an existing directory, or omit --evidence-dir", { verb: "run", flag: "evidence-dir" }, pretty);
+  }
   if (args.evidence) {
     try { bundle = readEvidence(args.evidence); } catch (e) {
       return emitError(`run: failed to read evidence bundle: ${e.message}`, { evidence: args.evidence }, pretty);
@@ -4176,11 +4208,12 @@ function cmdRunMulti(runner, ids, args, runOpts, pretty, meta) {
   // --evidence-dir <dir>: each <playbook-id>.json under the directory is read
   // as that playbook's submission. Lets operators wire up one cron job that
   // collects per-playbook evidence into a directory, then runs the whole
-  // contract in one pass.
+  // contract in one pass. The empty-string form is already refused above; the
+  // truthy gate here only ever sees a non-empty directory path.
   if (args["evidence-dir"]) {
     const dir = args["evidence-dir"];
-    if (typeof dir !== "string" || dir.length === 0) {
-      return emitError("run: --evidence-dir must be a non-empty string.", null, pretty);
+    if (typeof dir !== "string") {
+      return emitError("run: --evidence-dir must be a string.", null, pretty);
     }
     if (!fs.existsSync(dir)) {
       return emitError(`run: --evidence-dir ${dir} does not exist.`, null, pretty);
@@ -4951,7 +4984,11 @@ function walkAttestationDir(root, opts, candidates) {
         // Gate on the parsed kind so a renamed file cannot smuggle a replay
         // record into the listing.
         if (j && j.kind === "replay") continue;
-        if (opts.playbookId && j.playbook_id !== opts.playbookId) continue;
+        // Filter on an explicitly-supplied playbook id. `!= null` (not a
+        // truthiness check) so a future caller that threads an empty-string
+        // id can't silently disable the filter and widen the match to every
+        // playbook; the only legitimate "no filter" is `null`/`undefined`.
+        if (opts.playbookId != null && j.playbook_id !== opts.playbookId) continue;
         if (opts.since && (j.captured_at || "") < opts.since) continue;
         if (opts.excludeSessionId && sid === opts.excludeSessionId) continue;
         candidates.push({ sessionId: sid, playbookId: j.playbook_id, file: p, parsed: j });
@@ -5154,12 +5191,29 @@ function cmdPruneAttestations(runner, args, runOpts, pretty) {
       pretty,
     );
   }
-  const isoErr = validateIsoSince(cutoffRaw);
+  const isoErr = validateIsoSince(cutoffRaw, "--all-older-than");
   if (isoErr) return emitError(`attest prune: ${isoErr}`, { verb: "attest prune" }, pretty);
   const cutoffMs = Date.parse(cutoffRaw);
   const dryRun = !!args["dry-run"];
 
-  const roots = [...new Set([resolveAttestationRoot(runOpts), path.join(process.cwd(), ".exceptd", "attestations")])];
+  // Canonicalize before dedup. A plain Set over the two root strings only
+  // collapses byte-identical paths, so when the default root and the cwd-
+  // relative root resolve to the SAME directory via different strings (e.g. a
+  // relative EXCEPTD_HOME like `.exceptd`, or the home-mkdir-fail fallback),
+  // both survive and every session under that dir is scanned twice — inflating
+  // scanned/kept/pruned_count and double-listing each session in the preview.
+  // realpathSync resolves symlinks + makes absolute for an existing dir; for a
+  // not-yet-created root it throws, so fall back to path.resolve (absolute +
+  // normalized). Mirrors the realpath confinement used at delete time below.
+  const canonicalRoot = (p) => { try { return fs.realpathSync(p); } catch { return path.resolve(p); } };
+  const roots = [];
+  const seenRoots = new Set();
+  for (const r of [resolveAttestationRoot(runOpts), path.join(process.cwd(), ".exceptd", "attestations")]) {
+    const c = canonicalRoot(r);
+    if (seenRoots.has(c)) continue;
+    seenRoots.add(c);
+    roots.push(c);
+  }
   const pruned = [];
   let kept = 0;
   let scanned = 0;
@@ -5197,16 +5251,29 @@ function cmdPruneAttestations(runner, args, runOpts, pretty) {
       const ts = dateStr ? Date.parse(dateStr) : NaN;
       if (!Number.isFinite(ts)) { kept++; continue; }
       if (ts < cutoffMs) {
-        pruned.push({ session_id: sid, captured_at: captured, replayed_at: captured ? undefined : replayFallback, dir: sdir });
+        // Confinement: resolve and confirm sdir is a direct child of root
+        // before it can be deleted, so a crafted session name can't escape the
+        // root. Evaluate this in BOTH modes so the dry-run preview lists exactly
+        // the set a real run will remove — a session the real run would refuse
+        // (realpath escapes the root, or realpathSync throws) must not show up
+        // as [would-delete]. realDir is reused for the rmSync below so the
+        // delete and the gate operate on the same canonical path (no TOCTOU
+        // between the check and the removal).
+        let realDir = null;
+        try {
+          const realRoot = fs.realpathSync(root);
+          const candidate = fs.realpathSync(sdir);
+          if (path.dirname(candidate) === realRoot) realDir = candidate;
+        } catch { /* unresolvable -> not deletable */ }
+        if (realDir === null) { kept++; continue; }
         if (!dryRun) {
-          // Confinement: resolve and confirm sdir is a direct child of root
-          // before removing, so a crafted session name can't escape the root.
-          try {
-            const realRoot = fs.realpathSync(root);
-            const realDir = fs.realpathSync(sdir);
-            if (path.dirname(realDir) === realRoot) fs.rmSync(realDir, { recursive: true, force: true });
-          } catch { /* skip undeletable */ }
+          // Real run: count the session as pruned only after the delete
+          // succeeds, so pruned_count is a post-condition (sessions actually
+          // removed from disk), never a candidate tally.
+          try { fs.rmSync(realDir, { recursive: true, force: true }); }
+          catch { kept++; continue; /* skip undeletable */ }
         }
+        pruned.push({ session_id: sid, captured_at: captured, replayed_at: captured ? undefined : replayFallback, dir: sdir });
       } else {
         kept++;
       }
@@ -5246,13 +5313,27 @@ function cmdReattest(runner, args, runOpts, pretty) {
     const sinceErr = validateIsoSince(args.since);
     if (sinceErr) return emitError(`reattest: ${sinceErr}`, null, pretty);
   }
+  // Normalize --playbook (registered `multi:`, so a single value arrives as a
+  // one-element array) and refuse an empty value. `--playbook ""` would
+  // otherwise unwrap to "" and slip past walkAttestationDir's truthy filter
+  // guard — silently widening --latest to the newest attestation across ALL
+  // playbooks rather than the requested one. Refuse explicitly, the same way
+  // --since refuses a malformed value above, so the operator sees the bad
+  // input instead of an unintended cross-playbook match.
+  let playbookFilter = null;
+  if (args.playbook != null) {
+    playbookFilter = Array.isArray(args.playbook) ? args.playbook[0] : args.playbook;
+    if (typeof playbookFilter !== "string" || playbookFilter === "") {
+      return emitError("reattest: --playbook was given an empty value. Pass a playbook id (e.g. --playbook kernel) or omit --playbook to match across all playbooks.", { verb: "reattest", flag: "playbook" }, pretty);
+    }
+  }
   // --latest [--playbook <id>] [--since <ISO>] — find prior attestation
   // without requiring the operator to know the session-id.
   let sessionId = args._[0];
   let attFile = null;
   if (!sessionId && args.latest) {
     const found = findLatestAttestation({
-      playbookId: args.playbook ? (Array.isArray(args.playbook) ? args.playbook[0] : args.playbook) : null,
+      playbookId: playbookFilter,
       since: args.since || null,
     });
     if (!found) return emitError("reattest: --latest found no matching attestations.", { filter: { playbook: args.playbook || null, since: args.since || null } }, pretty);
@@ -5765,6 +5846,22 @@ function cmdAttest(runner, args, runOpts, pretty) {
     // comparison. Without --against, replays current state against prior
     // session (= reattest). With --against, compares two sessions A vs B
     // by evidence_hash + artifact-level field diff.
+    //
+    // An empty `--against ""` / `--against=` parses to the empty string, which
+    // is falsy — without this guard it would skip the explicit two-session
+    // branch and silently fall through to the auto-prior path, comparing
+    // against a DIFFERENT baseline than the operator named (a `--against
+    // "$VAR"` that expanded to empty is the common footgun). REQUIRES_VALUE
+    // only catches the value-less `--against` (parsed as `true`), not the
+    // empty-string form. Refuse explicitly so the dropped target is signalled
+    // rather than swapped under the operator.
+    if (args.against === "") {
+      return emitError(
+        'attest diff: --against was given an empty value; pass a session-id, or omit --against to diff against the most-recent prior.',
+        { verb: "attest diff", flag: "against" },
+        pretty
+      );
+    }
     if (args.against) {
       // Validate the --against id with the same gate as the primary sid, so a
       // traversal/garbage value (`../../etc/passwd`) gets the explicit "invalid
@@ -5870,14 +5967,23 @@ function cmdAttest(runner, args, runOpts, pretty) {
         // counts. Pre-0.11.8 (self.submission||{}).artifacts was undefined
         // for flat submissions; the diff returned all zeros even when
         // artifacts were present in observations.
-        artifact_diff: diffArtifacts(
-          normalizedArtifacts(self.submission, runner, self.playbook_id),
-          normalizedArtifacts(other.submission, runner, other.playbook_id)
-        ),
-        signal_override_diff: diffSignalOverrides(
-          normalizedSignalOverrides(self.submission, runner, self.playbook_id),
-          normalizedSignalOverrides(other.submission, runner, other.playbook_id)
-        ),
+        // The catalog stub stands in for an empty side ONLY when BOTH sides
+        // are empty — substituting it for one empty side while the peer passes
+        // through its real keys manufactures phantom drift (every catalog id
+        // the populated side did not submit reads as added/changed).
+        ...(() => {
+          const bothEmpty = !submissionHasData(self.submission) && !submissionHasData(other.submission);
+          return {
+            artifact_diff: diffArtifacts(
+              normalizedArtifacts(self.submission, runner, self.playbook_id, bothEmpty),
+              normalizedArtifacts(other.submission, runner, other.playbook_id, bothEmpty)
+            ),
+            signal_override_diff: diffSignalOverrides(
+              normalizedSignalOverrides(self.submission, runner, self.playbook_id, bothEmpty),
+              normalizedSignalOverrides(other.submission, runner, other.playbook_id, bothEmpty)
+            ),
+          };
+        })(),
       }, pretty, renderAttestDiff);
       return;
     }
@@ -5954,14 +6060,22 @@ function cmdAttest(runner, args, runOpts, pretty) {
       sidecar_verify: aSidecarVerify,
       a_sidecar_verify: aSidecarVerify,
       b_sidecar_verify: bSidecarVerify,
-      artifact_diff: diffArtifacts(
-        normalizedArtifacts(self.submission, runner, self.playbook_id),
-        normalizedArtifacts(other.submission, runner, other.playbook_id),
-      ),
-      signal_override_diff: diffSignalOverrides(
-        normalizedSignalOverrides(self.submission, runner, self.playbook_id),
-        normalizedSignalOverrides(other.submission, runner, other.playbook_id),
-      ),
+      // Catalog stub stands in for an empty side only when BOTH sides are
+      // empty (same peer-symmetric gate as the --against branch); real-vs-empty
+      // diffs the populated side's keys against {}, not against the full catalog.
+      ...(() => {
+        const bothEmpty = !submissionHasData(self.submission) && !submissionHasData(other.submission);
+        return {
+          artifact_diff: diffArtifacts(
+            normalizedArtifacts(self.submission, runner, self.playbook_id, bothEmpty),
+            normalizedArtifacts(other.submission, runner, other.playbook_id, bothEmpty),
+          ),
+          signal_override_diff: diffSignalOverrides(
+            normalizedSignalOverrides(self.submission, runner, self.playbook_id, bothEmpty),
+            normalizedSignalOverrides(other.submission, runner, other.playbook_id, bothEmpty),
+          ),
+        };
+      })(),
     }, pretty, renderAttestDiff);
     return;
   }
@@ -6232,9 +6346,29 @@ function cmdAttest(runner, args, runOpts, pretty) {
       run_opts: a.run_opts,
       artifacts_redacted: Object.fromEntries(Object.entries((a.submission && a.submission.artifacts) || {})
         .map(([k, v]) => [k, { captured: !!v.captured, reason: v.reason || null, redacted_value: "[redacted]" }])),
-      signal_overrides: (a.submission && a.submission.signal_overrides) || {},
+      // signal_overrides are operator-controllable: the contract canonicalizes
+      // hit/miss/inconclusive verdicts but does NOT reject free-form values
+      // (an unrecognized value surfaces a signal_override_unrecognized runtime
+      // error yet is still stored verbatim in the submission), and the sibling
+      // `<id>__fp_checks` keys carry arbitrary operator attestation maps. Under
+      // a bundle labelled "redacted ... suitable for audit submission" the only
+      // audit-meaningful content is the indicator verdict itself, so keep an
+      // exact hit/miss/inconclusive value and replace everything else (free-form
+      // strings, captured-data values, __fp_checks objects) with "[redacted]".
+      // Apply the same keyname denylist as signals_redacted so an obviously-
+      // sensitive key can't ride through under this field either. Matches the
+      // signal-value-redaction contract asserted in tests/cli-coverage.js.
+      signal_overrides: Object.fromEntries(Object.entries((a.submission && a.submission.signal_overrides) || {})
+        .filter(([k]) => !/_filter$|_key$|token|secret|password/i.test(k))
+        .map(([k, v]) => [k, (v === "hit" || v === "miss" || v === "inconclusive") ? v : "[redacted]"])),
+      // Redact the VALUES, not just drop obviously-sensitive keys: a submitted
+      // signal value can hold operator data (e.g. a captured credential string),
+      // and this field is labelled "redacted". The keyname denylist still drops
+      // the obviously-sensitive keys entirely; every retained key keeps only a
+      // "[redacted]" placeholder value, matching artifacts_redacted above.
       signals_redacted: Object.fromEntries(Object.entries((a.submission && a.submission.signals) || {})
-        .filter(([k]) => !/_filter$|_key$|token|secret|password/i.test(k))),
+        .filter(([k]) => !/_filter$|_key$|token|secret|password/i.test(k))
+        .map(([k]) => [k, "[redacted]"])),
       precondition_checks: (a.submission && a.submission.precondition_checks) || {},
     }));
 
@@ -6256,7 +6390,7 @@ function cmdAttest(runner, args, runOpts, pretty) {
         verb: "attest export",
         session_id: sessionId,
         exported_at: new Date().toISOString(),
-        redaction_policy: "v0.10.3-default — artifact values stripped; signal_overrides + precondition_checks + evidence_hash + signature preserved.",
+        redaction_policy: "v0.10.3-default — artifact values stripped; signal_overrides reduced to hit/miss/inconclusive verdicts (free-form values redacted); precondition_checks + evidence_hash + signature preserved.",
         attestations: redacted,
       }, pretty);
     }
@@ -6296,9 +6430,25 @@ function _playbookSignalCatalog(runner, playbookId) {
     return Object.fromEntries(inds.map(i => [i.id, 'inconclusive']));
   } catch { return null; }
 }
-function normalizedArtifacts(submission, runner, playbookId) {
+// A submission carries real operator data for a diff when it supplied
+// artifacts, signal_overrides, OR observations. The empty case ({} or
+// {observations:{}}) is "no operator data was supplied." Diff symmetry hinges
+// on this predicate: the playbook catalog stub may only stand in for an empty
+// side when BOTH sides are empty (so the count reflects "N catalog ids,
+// uniformly empty on both sides"). Substituting the full catalog for one empty
+// side while the peer passes through its real keys manufactures phantom drift —
+// every catalog id the populated side did not submit shows up as "added"
+// (artifacts) or "changed" (signals). See callers for the bothEmpty gate.
+function submissionHasData(submission) {
+  if (!submission || typeof submission !== "object") return false;
+  const nonEmpty = (o) => o && typeof o === "object" && Object.keys(o).length > 0;
+  return nonEmpty(submission.artifacts)
+    || nonEmpty(submission.signal_overrides)
+    || nonEmpty(submission.observations);
+}
+function normalizedArtifacts(submission, runner, playbookId, applyEmptyFallback = true) {
   if (!submission || typeof submission !== "object") {
-    return _playbookArtifactCatalog(runner, playbookId) || {};
+    return applyEmptyFallback ? (_playbookArtifactCatalog(runner, playbookId) || {}) : {};
   }
   if (submission.artifacts && Object.keys(submission.artifacts).length > 0) return submission.artifacts;
   if (submission.observations && Object.keys(submission.observations).length > 0) {
@@ -6320,15 +6470,16 @@ function normalizedArtifacts(submission, runner, playbookId) {
     }
     return out;
   }
-  // v0.11.13 (#128): empty submission ({} or {observations:{}}). Identical
-  // hashes still mean "no operator data was supplied, same on both sides."
-  // Fall back to the playbook's look.artifacts catalog so total_compared
-  // reflects "N catalog artifacts, all uniformly empty on both sides."
-  return _playbookArtifactCatalog(runner, playbookId) || {};
+  // Empty submission ({} or {observations:{}}). The catalog stub may only
+  // stand in when the PEER side is also empty (applyEmptyFallback). When the
+  // peer carried real artifacts, return an empty map so the populated side's
+  // keys diff against nothing — yielding genuine added/removed instead of one
+  // fabricated "added" per catalog id the operator never submitted.
+  return applyEmptyFallback ? (_playbookArtifactCatalog(runner, playbookId) || {}) : {};
 }
-function normalizedSignalOverrides(submission, runner, playbookId) {
+function normalizedSignalOverrides(submission, runner, playbookId, applyEmptyFallback = true) {
   if (!submission || typeof submission !== "object") {
-    return _playbookSignalCatalog(runner, playbookId) || {};
+    return applyEmptyFallback ? (_playbookSignalCatalog(runner, playbookId) || {}) : {};
   }
   if (submission.signal_overrides && Object.keys(submission.signal_overrides).length > 0) return submission.signal_overrides;
   if (submission.observations && Object.keys(submission.observations).length > 0) {
@@ -6347,7 +6498,10 @@ function normalizedSignalOverrides(submission, runner, playbookId) {
     }
     return out;
   }
-  return _playbookSignalCatalog(runner, playbookId) || {};
+  // Empty submission — same peer-symmetric gate as normalizedArtifacts: only
+  // stand in the inconclusive catalog stub when the peer is also empty, so a
+  // real-vs-empty signal diff reports only the genuinely-differing indicators.
+  return applyEmptyFallback ? (_playbookSignalCatalog(runner, playbookId) || {}) : {};
 }
 
 /**
@@ -6398,14 +6552,14 @@ function diffArtifacts(a, b) {
   for (const id of allIds) {
     const av = a[id], bv = b[id];
     if (!av && bv) {
-      out.added.push({ id, captured: !!bv.captured, value_preview: previewValue(bv.value) });
+      out.added.push({ id, captured: !!bv.captured, value_preview: artifactPreview(bv) });
     } else if (av && !bv) {
-      out.removed.push({ id, captured: !!av.captured, value_preview: previewValue(av.value) });
+      out.removed.push({ id, captured: !!av.captured, value_preview: artifactPreview(av) });
     } else if (av && bv && artifactsDiffer(av, bv)) {
       out.changed.push({
         id,
         a_captured: !!av.captured, b_captured: !!bv.captured,
-        a_value_preview: previewValue(av.value), b_value_preview: previewValue(bv.value),
+        a_value_preview: artifactPreview(av), b_value_preview: artifactPreview(bv),
       });
     } else if (av && bv) {
       // v0.11.8 (#102): both sides have the entry AND they're identical →
@@ -6438,6 +6592,23 @@ function previewValue(v) {
   return s.length > 80 ? s.slice(0, 80) + "…" : s;
 }
 
+// Preview the evidence an artifact carries for the diff output. `.value` is the
+// canonical carrier, but observations legitimately store their secret/path/match
+// under other keys (path, matched, reason, or a custom key). When `.value` is
+// absent, fall back to a preview of the remaining evidence-bearing keys (every
+// key except the bookkeeping `captured`/`captured_at` flags) so non-`value`
+// carriers still render instead of collapsing to a null preview — which hid the
+// actual differing content even when the per-field equality compare correctly
+// flagged the artifact as changed.
+function artifactPreview(art) {
+  if (art === null || typeof art !== "object" || Array.isArray(art)) return previewValue(art);
+  if (art.value !== undefined && art.value !== null) return previewValue(art.value);
+  const { captured, captured_at, _captured_at, value, ...evidence } = art;
+  const keys = Object.keys(evidence);
+  if (keys.length === 0) return null;
+  return previewValue(evidence);
+}
+
 // ---------------------------------------------------------------------------
 // v0.11.0: cmdDiscover — context-aware playbook recommender.
 // Collapses scan + dispatch + recommend into one verb. Sniffs the cwd, reads
@@ -6448,6 +6619,11 @@ function cmdDiscover(runner, args, runOpts, pretty) {
   // process cwd. Pre-fix it was silently ignored — recommendations were
   // computed for the wrong directory with no signal. Validated like collect.
   let cwd = process.cwd();
+  // An explicit empty value (`--cwd ""`) would otherwise be falsy and silently
+  // scan process.cwd() — the wrong directory — reported as a successful run.
+  if (args.cwd === "") {
+    return emitError(`discover: --cwd was given an empty value; pass an existing directory path`, { verb: "discover" }, pretty);
+  }
   if (args.cwd) {
     const resolved = path.resolve(String(args.cwd));
     let stat;
@@ -7935,6 +8111,14 @@ function cmdAiRun(runner, args, runOpts, pretty) {
   if (!playbookId) {
     return emitError("ai-run: missing <playbook> positional argument.", null, pretty);
   }
+  // An explicit empty value (`--evidence ""`) is operator error, same as `run`.
+  // The `--no-stream` path tests `args.evidence` for truthiness, so `""` fell
+  // through to the stdin branch and — with empty/closed stdin — ran an empty
+  // submission to ok:true at exit 0, masking that the intended evidence never
+  // loaded. Reject it here so both stream and no-stream entry behave like `run`.
+  if (args.evidence === "") {
+    return emitError("ai-run: --evidence was given an empty value; pass a file path, '-' for stdin, or omit --evidence to read evidence from the stream", { verb: "ai-run" }, pretty);
+  }
   if (refuseInvalidPlaybookId("ai-run", playbookId, pretty)) return;
   let pb;
   try { pb = runner.loadPlaybook(playbookId); }
@@ -8684,6 +8868,27 @@ function cmdCi(runner, args, runOpts, pretty) {
       pretty,
     );
   }
+  // An explicit empty value (`--evidence ""` / `--evidence=` / an unset shell
+  // variable) is falsy, so the truthiness-gated evidence reads below skip
+  // entirely, the bundle stays {}, every playbook runs with no evidence, and
+  // the gate reports a clean PASS at exit 0 — a false-green that hides the fact
+  // the operator's intended evidence never loaded. Mirror the run/collect/
+  // discover empty-value guards: refuse the empty value loudly rather than
+  // running a vacuous gate.
+  if (args.evidence === "") {
+    return emitError(
+      "ci: --evidence was given an empty value; pass a file path, '-' for stdin, or omit --evidence for a no-evidence run",
+      { verb: "ci", flag: "evidence" },
+      pretty,
+    );
+  }
+  if (args["evidence-dir"] === "") {
+    return emitError(
+      "ci: --evidence-dir was given an empty value; pass an existing directory, or omit --evidence-dir",
+      { verb: "ci", flag: "evidence-dir" },
+      pretty,
+    );
+  }
   const blockOnClock = !!args["block-on-jurisdiction-clock"];
 
   // v0.11.9 (#115): --required <playbook,playbook,...> takes precedence over

package/data/_indexes/_meta.json

@@ -1,14 +1,14 @@
 {
   "schema_version": "1.1.0",
-  "generated_at": "2026-06-14T16:18:12.210Z",
+  "generated_at": "2026-06-20T16:35:33.578Z",
   "generator": "scripts/build-indexes.js",
   "source_count": 64,
   "source_hashes": {
-    "manifest.json": "2ddf8b26d75612209fb45014014c2729b0d6228e8a0eb3b6c2fb90b5997ce3f5",
+    "manifest.json": "6e8a5ac562c706b0142ad12b07a62982bb83d500870ca33a5d865bd17a16d23a",
     "README.md": "e7b854e7db9a364a1b368b5084b4f0c2a8282f0459ce39800ac1d1dabdc06074",
     "data/atlas-ttps.json": "5bc59e23d6c2defa54168de161a0825299b9cc4a49c6b26df2dae70b4f42eedf",
     "data/attack-techniques.json": "53c6f248760eecb11a0354f74ab467a5814e95075a686b9b3bf18c34e2f7435e",
-    "data/cve-catalog.json": "eedd129c657e535d53df5d4d44fa54009b69cde85673cf7c9d426054eddc43d2",
+    "data/cve-catalog.json": "06ca53e69071dfe94867c10717f3ff2962c50341defcaf050f1a96236b5be51c",
     "data/cwe-catalog.json": "359263361fa52069e2856cc352d8f1c757d614ea840db6ef3ff5e696185ca220",
     "data/d3fend-catalog.json": "349d14b2777342d38e5f8b0149a9ebd7703ee59f67a9efee200d1e19313088ac",
     "data/dlp-controls.json": "d2406c482dddd30e49203879999dc4b3a7fd4d0494d6a61d86b91ee76415df19",
@@ -71,6 +71,25 @@
   },
   "skill_count": 51,
   "catalog_count": 11,
+  "outputs": [
+    "activity-feed.json",
+    "catalog-summaries.json",
+    "chains.json",
+    "currency.json",
+    "did-ladders.json",
+    "frequency.json",
+    "handoff-dag.json",
+    "jurisdiction-clocks.json",
+    "jurisdiction-map.json",
+    "recipes.json",
+    "section-offsets.json",
+    "stale-content.json",
+    "summary-cards.json",
+    "theater-fingerprints.json",
+    "token-budget.json",
+    "trigger-table.json",
+    "xref.json"
+  ],
   "index_stats": {
     "xref_entries": {
       "cwe_refs": 53,