mjolnir-qa 2.1.0 → 3.0.0

package/dist/cli.mjs

@@ -1,10 +1,10 @@
 #!/usr/bin/env node
-import { $ as massCeiling, A as box, At as isAdvisoryFinding, B as headlineFor, C as buildFooter, Ct as buildEvidenceGraph, D as plainContext, Dt as RULE_CATEGORIES, E as panel, Et as QA_IMPACT_LABELS, F as scoreGauge, G as buildDependencyGraph, H as EVIDENCE, I as shouldColorize, J as classifyProvenance, K as getReachableFiles, L as shouldUseAscii, M as padTo, N as palette, O as sectionHeader, Ot as SEVERITY_ORDER$1, P as sanitizeData, Q as deductionFor, R as wrapText, S as FLAKE_GLYPH, St as ENGINE_VERSION, T as okIcon, Tt as DEDUCTIONS, U as TINT, V as BADGE_BAND, W as TRUST, X as capForTier, Y as computeAgenticProfile, Z as computeDimensions, _t as ConfigValidationError, at as SEARCHED_FOR, b as renderSuppressions, bt as isRecord$2, ct as parseWorkflow, dt as parseTsFile, et as RETIRED_RULE_IDS, ft as detectFrameworks, gt as isLintFixtureDir, h as scan_pipeline_exports, ht as globToRegExp, it as SCAN_ADAPTERS, j as measure, jt as isValidCategory, k as severityIcon, kt as deriveEvidenceLevel, lt as parseYamlGuarded, m as runScan$1, mt as createIgnoreMatcher, n as KNOWN_RULE_IDS$1, nt as getRule, ot as isAzurePipelineFixture, pt as DEFAULT_IGNORE_MATCHER, q as correlateFindings, rt as resolveGitPath, st as parseAzurePipeline, tt as RULES, ut as computeCodeText, v as loadLocalRules, vt as isSuppressionActive, w as nextStep, wt as MEASURED_FP, x as runForensics, xt as parseJsonFile, y as loadSuppressions, yt as loadConfig, z as deriveScoreState } from "./scan-pipeline-MQGUSey3.mjs";
-import { accessSync, appendFileSync, chmodSync, closeSync, constants, copyFileSync, existsSync, lstatSync, mkdirSync, mkdtempSync, openSync, readFileSync, readdirSync, realpathSync, renameSync, rmSync, statSync, unlinkSync, writeFileSync, writeSync } from "node:fs";
+import { $ as correlateFindings, A as plainContext, At as parseJsonFile, B as shouldUseAscii, Bt as isValidCategory, C as runForensics, Ct as globToRegExp, D as nextStep, Dt as loadConfig, E as buildFooter, Et as isSuppressionActive, F as padTo, Ft as QA_IMPACT_LABELS, G as EVIDENCE, H as deriveScoreState, I as palette, It as RULE_CATEGORIES, J as atomicTempPath, K as TINT, L as sanitizeData, Lt as SEVERITY_ORDER$1, M as severityIcon, Mt as buildEvidenceGraph, N as box, Nt as MEASURED_FP, O as okIcon, Ot as readFileBounded, P as measure, Pt as DEDUCTIONS, Q as getReachableFiles, R as scoreGauge, Rt as deriveEvidenceLevel, S as renderSuppressions, St as createIgnoreMatcher, T as FLAKE_GLYPH, Tt as ConfigValidationError, U as headlineFor, V as wrapText, W as BADGE_BAND, X as writeFileAtomic, Y as sweepStaleTempFiles, Z as buildDependencyGraph, _t as computeCodeText, at as massCeiling, b as loadLocalRules, bt as DEFAULT_IGNORE_MATCHER, ct as getRule, dt as SCAN_ADAPTERS, et as classifyProvenance, ft as SEARCHED_FOR, gt as parseYamlGuarded, h as scan_pipeline_exports, ht as parseWorkflow, it as deductionFor, j as sectionHeader, jt as ENGINE_VERSION, k as panel, kt as isRecord$3, lt as resolveGitPath, m as runScan$1, mt as parseAzurePipeline, n as KNOWN_RULE_IDS$1, nt as capForTier, ot as RETIRED_RULE_IDS, pt as isAzurePipelineFixture, q as TRUST, rt as computeDimensions, st as RULES, tt as computeAgenticProfile, ut as runGit, v as pluginsGateOpen, vt as parseTsFile, w as sanitizeErrorText, wt as isLintFixtureDir, x as loadSuppressions, xt as LIMITS, y as renderGateNotice, yt as detectFrameworks, z as shouldColorize, zt as isAdvisoryFinding } from "./scan-pipeline-CAH9_Qgh.mjs";
+import { accessSync, appendFileSync, chmodSync, constants, copyFileSync, existsSync, lstatSync, mkdirSync, mkdtempSync, readFileSync, readdirSync, realpathSync, renameSync, rmSync, statSync, unlinkSync, writeFileSync } from "node:fs";
 import { basename, dirname, isAbsolute, join, relative, resolve, sep } from "node:path";
 import process$1 from "node:process";
 import { fileURLToPath, pathToFileURL } from "node:url";
-import { createHash, randomBytes } from "node:crypto";
+import { createHash } from "node:crypto";
 import ts, { ts as ts$1 } from "ts-morph";
 import { parse } from "yaml";
 import { execFileSync } from "node:child_process";
@@ -672,27 +672,21 @@ concurrency:
 
 permissions:
   contents: read
-  pull-requests: write
 
 jobs:
   scan:
     runs-on: ubuntu-latest
-    # A hung scan must not sit for the 6-hour default.
     timeout-minutes: 10
+    permissions:
+      contents: read
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
         with:
-          fetch-depth: 0   # needed for --scope changed merge-base
+          fetch-depth: 0
           persist-credentials: false
-      # Scan with the pinned version. No subcommand: npx resolves
-      # the package bin directly; passing mjolnir as an arg causes
-      # npm 10+ npx to invoke it twice (unknown subcommand).
       - name: Scan changed code (exit 1/2 is data — the gate step decides)
         continue-on-error: true
         run: npx --yes https://registry.npmjs.org/mjolnir-qa/-/mjolnir-qa-${ENGINE_VERSION}.tgz . --scope changed --json > mjolnir.json
-      # Reporting, not gating: a crashed scan leaves mjolnir.json empty/missing
-      # and the summary step exits 2/10 — continue-on-error keeps the advisory
-      # job green, exactly like the v1 inline script did (the gate step decides).
       - name: Annotations + Job Summary
         if: always()
         continue-on-error: true
@@ -700,32 +694,49 @@ jobs:
       - name: Render PR comment
         if: always()
         continue-on-error: true
-        run: npx --yes https://registry.npmjs.org/mjolnir-qa/-/mjolnir-qa-${ENGINE_VERSION}.tgz pr-comment . > mjolnir-comment.md
-      # Best-effort: on a pull_request event from a fork the GITHUB_TOKEN is
-      # read-only and this step will 403 for every external contributor. The
-      # Job Summary above is the fallback that always renders.
-      # (pull_request_target would fix the token but is a code-execution
-      # risk — deliberately NOT used.)
-      - name: Post or update PR comment
+        run: npx --yes https://registry.npmjs.org/mjolnir-qa/-/mjolnir-qa-${ENGINE_VERSION}.tgz pr-comment --from mjolnir.json > mjolnir-comment.md
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        if: always()
+        with:
+          name: mjolnir-report
+          path: mjolnir-comment.md
+          if-no-files-found: warn
+          retention-days: 30
+${gate === "advisory" ? `      - name: Gate (advisory)
+        if: always()
+        run: echo "Advisory mode — findings reported, never blocking."` : `      - name: Gate (${gate})
+        if: always()
+        run: |
+          node -e '
+${indentBlock(gateScript(gate), 10)}
+          '`}
+
+  publish:
+    needs: scan
+    if: always() && needs.scan.result != 'cancelled'
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    permissions:
+      contents: read
+      pull-requests: write
+    steps:
+      - uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        with:
+          name: mjolnir-report
+      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         if: always()
         continue-on-error: true
-        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         with:
           script: |
             const fs = require('fs');
-            let body = '';
-            try { body = fs.readFileSync('mjolnir-comment.md', 'utf8'); } catch (e) {}
-            if (!body.trim()) {
-              console.log('mjolnir-comment.md is empty or missing — nothing to post.');
-              return;
-            }
-            const marker = '<!-- mjolnir-pr-comment -->';
-            const { data: comments } = await github.rest.issues.listComments({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              issue_number: context.issue.number,
-            });
-            const existing = comments.find((c) => c.body?.startsWith(marker));
+            const body = fs.readFileSync('mjolnir-comment.md', 'utf8');
+            if (!body.trim()) return;
+            const marker = '<!-- mjolnir-report:v2 -->';
+            const comments = await github.paginate(
+              github.rest.issues.listComments,
+              { owner: context.repo.owner, repo: context.repo.repo, issue_number: context.issue.number, per_page: 100 },
+            );
+            const existing = comments.find((comment) => comment.body?.includes(marker));
             if (existing) {
               await github.rest.issues.updateComment({
                 owner: context.repo.owner,
@@ -741,20 +752,9 @@ jobs:
                 body,
               });
             }
-${gate === "advisory" ? `      # Advisory mode: findings are reported in the Job Summary and the
-      # PR comment, never blocking. The scan step's exit code is visible
-      # as the step outcome, but continue-on-error keeps the job green.
-      - name: Gate (advisory)
-        if: always()
-        run: echo "Advisory mode — findings reported, never blocking."` : `      # Gate enforcement: the scan step's own exit code is deliberately
-      # neutralized (continue-on-error) so reporting steps always run; THIS
-      # step is what fails the job. A partial scan never blocks.
-      - name: Gate (${gate})
+${gate === "advisory" ? `      - name: Advisory note
         if: always()
-        run: |
-          node -e '
-${indentBlock(gateScript(gate), 10)}
-          '`}
+        run: echo "Advisory mode — findings are reported, never blocking."` : ""}
 `;
 /** Every gate variant, used to tell "our template" from "hand-customized". */
 const GATES = [
@@ -793,69 +793,77 @@ concurrency:
 
 permissions:
   contents: read
-  pull-requests: write
 
 jobs:
   scan:
     runs-on: ubuntu-latest
-    # A hung scan must not sit for the 6-hour default.
     timeout-minutes: 10
+    permissions:
+      contents: read
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
         with:
-          fetch-depth: 0   # needed for --scope changed merge-base
+          fetch-depth: 0
           persist-credentials: false
-      # The action scans with the published mjolnir-qa package
-      # pinned to this CLI release. The composite action is pinned via
-      # ACTION_REF. The fail-on input is the gate: it fails the
-      # job on findings at the gate and never on a partial scan
-      # (exit 2 downgrades to a warning — the frozen exit-code
-      # contract). Advisory mode reports, never blocks.
       - name: Mjölnir verification trust scan
         id: mjolnir
-        if: always()
-        continue-on-error: ${gate === "advisory" ? "true" : "false"}
+        continue-on-error: true
         uses: ${ACTION_REF}
         with:
           scope: changed
           format: json
           fail-on: ${gate === "advisory" ? "none" : gate}
           version: ${ENGINE_VERSION}
-      # Reporting, not gating: runs even when the scan/gate failed, from
-      # the same mjolnir.json the action wrote.
-      - name: Annotations + Job Summary
+          pr-comment: "false"
+          trust-artifact: "false"
+          annotations: "true"
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         if: always()
-        continue-on-error: true
-        run: npx --yes https://registry.npmjs.org/mjolnir-qa/-/mjolnir-qa-${ENGINE_VERSION}.tgz summary mjolnir.json
-      - name: Render PR comment
+        with:
+          name: mjolnir-report
+          path: mjolnir-comment.md
+          if-no-files-found: warn
+          retention-days: 30
+      - name: Gate (${gate})
         if: always()
-        continue-on-error: true
-        run: npx --yes https://registry.npmjs.org/mjolnir-qa/-/mjolnir-qa-${ENGINE_VERSION}.tgz pr-comment . > mjolnir-comment.md
-      # Best-effort: on a pull_request event from a fork the GITHUB_TOKEN is
-      # read-only and this step will 403 for every external contributor. The
-      # Job Summary above is the fallback that always renders.
-      # (pull_request_target would fix the token but is a code-execution
-      # risk — deliberately NOT used.)
-      - name: Post or update PR comment
+        shell: bash
+        env:
+          MJ_SCAN_EXIT: \${{ steps.mjolnir.outputs.exit }}
+        run: |
+          if [ "$MJ_SCAN_EXIT" = "1" ]; then
+            exit 1
+          fi
+          if [ "$MJ_SCAN_EXIT" = "2" ]; then
+            echo "::warning::Mjolnir analysis is partial; reporting remains visible but the gate is not claimed."
+          fi
+          exit 0
+
+  publish:
+    needs: scan
+    if: always() && needs.scan.result != 'cancelled'
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    permissions:
+      contents: read
+      pull-requests: write
+    steps:
+      - uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        with:
+          name: mjolnir-report
+      - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         if: always()
         continue-on-error: true
-        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         with:
           script: |
             const fs = require('fs');
-            let body = '';
-            try { body = fs.readFileSync('mjolnir-comment.md', 'utf8'); } catch (e) {}
-            if (!body.trim()) {
-              console.log('mjolnir-comment.md is empty or missing — nothing to post.');
-              return;
-            }
-            const marker = '<!-- mjolnir-pr-comment -->';
-            const listed = await github.rest.issues.listComments({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              issue_number: context.issue.number,
-            });
-            const existing = listed.data.find((c) => c.body?.startsWith(marker));
+            const body = fs.readFileSync('mjolnir-comment.md', 'utf8');
+            if (!body.trim()) return;
+            const marker = '<!-- mjolnir-report:v2 -->';
+            const comments = await github.paginate(
+              github.rest.issues.listComments,
+              { owner: context.repo.owner, repo: context.repo.repo, issue_number: context.issue.number, per_page: 100 },
+            );
+            const existing = comments.find((comment) => comment.body?.includes(marker));
             if (existing) {
               await github.rest.issues.updateComment({
                 owner: context.repo.owner,
@@ -871,17 +879,9 @@ jobs:
                 body,
               });
             }
-${gate === "advisory" ? `      # Advisory mode: findings are reported in the Job Summary,
-      # never blocking — the action ran with fail-on: none and
-      # continue-on-error, so even a crashed scan cannot fail this job.
-      - name: Gate (advisory)
+${gate === "advisory" ? `      - name: Advisory note
         if: always()
-        run: echo "Advisory mode — findings reported, never blocking."` : `      # Gate enforcement: the action step above IS the gate — fail-on
-      # ${gate} exits 1 on findings at the gate and the step is
-      # continue-on-error: false, so its failure fails the job. A partial
-      # scan never blocks (the action downgrades exit 2 to a warning per
-      # the frozen exit-code contract). The reporting steps above ran
-      # first (if: always()), so a red gate never suppresses the report.`}
+        run: echo "Advisory mode — findings are reported, never blocking."` : ""}
 `;
 /** Multiset line diff — counts only, for the refusal message. */
 function summarizeContentDiff(existing, incoming) {
@@ -940,6 +940,10 @@ function digestView(f) {
 		trustLevel: f.trustLevel ?? null,
 		confidence: f.confidence,
 		findingType: f.findingType,
+		qaImpact: f.qaImpact,
+		findingId: f.findingId ?? null,
+		rootCauseId: f.rootCauseId ?? null,
+		deduplicationGroup: f.deduplicationGroup ?? null,
 		fixGroupId: f.fixGroupId ?? null
 	};
 }
@@ -960,8 +964,12 @@ function canonicalScanJson(result) {
 			rules: result.analysisStatus.rules,
 			skippedFiles: result.analysisStatus.skippedFiles,
 			rulesCrashed: result.analysisStatus.rulesCrashed ?? 0,
-			truncationReasons: result.analysisStatus.truncationReasons ?? []
+			parseFallbacks: result.analysisStatus.parseFallbacks ?? 0,
+			truncationReasons: result.analysisStatus.truncationReasons ?? [],
+			reasons: result.analysisStatus.reasons ?? []
 		},
+		scopeIntegrity: result.scopeIntegrity ?? null,
+		runIdentity: result.runIdentity ?? null,
 		findings: result.findings.map(digestView)
 	});
 }
@@ -1018,9 +1026,13 @@ function buildMachineContract(result) {
 			rules: result.analysisStatus.rules,
 			skippedFiles: result.analysisStatus.skippedFiles,
 			rulesCrashed: result.analysisStatus.rulesCrashed ?? 0,
+			...result.analysisStatus.parseFallbacks !== void 0 ? { parseFallbacks: result.analysisStatus.parseFallbacks } : {},
 			truncationReasons: result.analysisStatus.truncationReasons ?? [],
+			...result.analysisStatus.reasons !== void 0 ? { reasons: result.analysisStatus.reasons } : {},
 			frameworkDetectionUnknown: result.frameworkDetectionUnknown,
-			durationMs: result.analysisStatus.durationMs
+			durationMs: result.analysisStatus.durationMs,
+			...result.scopeIntegrity !== void 0 ? { scopeIntegrity: result.scopeIntegrity } : {},
+			...result.runIdentity !== void 0 ? { runIdentity: result.runIdentity } : {}
 		},
 		...result.trustSummary !== void 0 ? { trustSummary: result.trustSummary } : {},
 		...result.agenticProfile !== void 0 ? { provenance: result.agenticProfile } : {},
@@ -1294,7 +1306,7 @@ function whatWouldChangeTheVerdict(r) {
 /** Structural validation of the loaded JSON (hostile-input safe). */
 function parseScanJson(raw) {
 	try {
-		return parseJsonFile(raw, "scan result", (v) => isRecord$2(v) && v.schemaVersion === 1 && Array.isArray(v.findings) && Boolean(v.analysisStatus) && typeof v.partial === "boolean");
+		return parseJsonFile(raw, "scan result", (v) => isRecord$3(v) && v.schemaVersion === 1 && Array.isArray(v.findings) && Boolean(v.analysisStatus) && typeof v.partial === "boolean");
 	} catch {
 		return;
 	}
@@ -1409,122 +1421,6 @@ function pctOf(v) {
 	return `${Math.round(v * 100)}%`;
 }
 //#endregion
-//#region src/lib/fs-atomic.ts
-/**
-* Atomic file writes (audit S9).
-*
-* Every durability-critical write in Mjölnir (baseline, stats, badge,
-* TRIAGE.md, scaffolded rule files) used to hand-roll
-* `writeFileSync(path, data)` — a crash mid-write left a TRUNCATED file
-* at the real path, and a subsequent read (diff, badge endpoint) served
-* confident nonsense from it.
-*
-* `writeFileAtomic` writes to a temp sibling, then RENAMES. On the same
-* volume rename is atomic: readers see either the complete old file or
-* the complete new file, never a half-written one. The temp name is
-* created with `wx` (exclusive) so concurrent writers cannot interleave,
-* stale temps are cleaned up on failure, and on Windows the rename is
-* retried briefly because a concurrent reader can hold the destination
-* open (EBUSY/EPERM).
-*/
-function atomicTempPath(path) {
-	return `${path}.mjolnir-${process.pid}-${Date.now()}-${randomBytes(4).toString("hex")}.tmp`;
-}
-/**
-* Atomically replace `path` with `data`.
-*/
-function writeFileAtomic(path, data, opts = {}) {
-	const dir = dirname(path);
-	if (opts.mkdirs !== false && !existsSync(dir)) mkdirSync(dir, { recursive: true });
-	const tmp = atomicTempPath(path);
-	let fd;
-	try {
-		fd = openSync(tmp, "wx", opts.mode ?? 420);
-		writeSync(fd, data, null, opts.encoding ?? "utf8");
-	} finally {
-		if (fd !== void 0) closeSync(fd);
-	}
-	try {
-		renameWithWindowsRetry(tmp, path);
-	} catch (err) {
-		try {
-			if (existsSync(tmp)) unlinkSync(tmp);
-		} catch {}
-		throw err;
-	}
-}
-/**
-* renameSync retry loop for Windows: a concurrent reader (another scan,
-* a badge endpoint, an editor) holding the destination open makes
-* rename fail with EBUSY/EPERM. A short bounded retry closes the race
-* without turning an atomic swap into a partial write.
-*
-* Internal contract test hook: the platform check keeps this loop off
-* the POSIX hot path; on win32 the EBUSY/EPERM arms are exercised by
-* the fs-atomic-retry spec (mocked renameSync).
-*/
-const RENAME_RETRIES = 8;
-const RENAME_RETRY_DELAY_MS = 25;
-/** Synchronous sleep that does not spin the CPU. */
-function sleepSync(ms) {
-	try {
-		Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0, ms);
-	} catch {
-		const until = Date.now() + ms;
-		while (Date.now() < until);
-	}
-}
-function renameWithWindowsRetry(from, to) {
-	for (let attempt = 0;; attempt++) try {
-		renameSync(from, to);
-		return;
-	} catch (err) {
-		const code = err?.code;
-		if (process.platform === "win32" && (code === "EBUSY" || code === "EPERM") && attempt < RENAME_RETRIES) {
-			sleepSync(RENAME_RETRY_DELAY_MS);
-			continue;
-		}
-		throw err;
-	}
-}
-const STALE_TEMP_AGE_MS = 864e5;
-const TEMP_PID_RE = /^.+\.mjolnir-([1-9]\d{0,9})-(\d{13})-[0-9a-f]{8}\.tmp$/;
-function sweepStaleTempFiles(dir) {
-	let swept = 0;
-	let entries;
-	try {
-		entries = readdirSync(dir);
-	} catch {
-		return 0;
-	}
-	const cutoff = Date.now() - STALE_TEMP_AGE_MS;
-	for (const entry of entries) {
-		const match = TEMP_PID_RE.exec(entry);
-		if (!match || Number(match[2]) > cutoff) continue;
-		const path = join(dir, entry);
-		try {
-			const stat = lstatSync(path);
-			if (!stat.isFile() || stat.mtimeMs > cutoff) continue;
-			if (pidAlive(Number(match[1]))) continue;
-			unlinkSync(path);
-			swept++;
-		} catch {
-			continue;
-		}
-	}
-	return swept;
-}
-function pidAlive(pid) {
-	if (!Number.isSafeInteger(pid) || pid <= 0 || pid > 2147483647) return true;
-	if (pid === process.pid) return true;
-	try {
-		process.kill(pid, 0);
-		return true;
-	} catch (err) {
-		return err?.code !== "ESRCH";
-	}
-}
-//#endregion
 //#region src/engine/finding-identity.ts
 /**
 * Finding Identity (ENGINE-002): deterministic, stable fingerprints for
@@ -2029,12 +1925,15 @@ function trustVerdictFor(v) {
 * today the honest command set is repeat/compare/inspect, not
 * "collect a trace", because trace.zip is not yet ingested.
 */
+function safeTriageText(value) {
+	return value.replace(/[;&|`$()\\\r\n]/g, " ").slice(0, 200);
+}
 function nextActionFor(v) {
 	switch (classifyVerdict(v)) {
-		case "RETRY-DEPENDENT": return `repeat execution of this test (3+ runs) to confirm the flake rate, then quarantine + ticket: ${v.file}`;
-		case "TIMEOUT": return `re-run in isolation to separate slowness from a hang: npx playwright test ${v.file} --timeout 60000`;
-		case "FAILING": return `reproduce locally: npx playwright test ${v.file} -g "${v.title.replace(/\\/g, "\\\\").replace(/"/g, "\\\"")}" — then fix and re-run`;
-		case "SKIPPED": return `inspect the skip condition in ${v.file} — a skip after a failure usually hides an environment problem`;
+		case "RETRY-DEPENDENT": return `repeat execution of this test (3+ runs) to confirm the flake rate, then quarantine + ticket: ${safeTriageText(v.file)}`;
+		case "TIMEOUT": return `re-run this test in isolation with a 60-second timeout to separate slowness from a hang: ${safeTriageText(v.file)}`;
+		case "FAILING": return `reproduce this test in isolation, then fix and re-run: ${safeTriageText(v.file)} — ${safeTriageText(v.title)}`;
+		case "SKIPPED": return `inspect the skip condition in ${safeTriageText(v.file)} — a skip after a failure usually hides an environment problem`;
 	}
 }
 /** Deterministic guided rows, worst first (same order law as triageRows). */
@@ -2068,7 +1967,7 @@ function renderTriageWorkflow(report) {
 	}
 	for (const r of rows) {
 		const flag = r.classification === "RETRY-DEPENDENT" ? "TRUE-FLAKE" : r.classification === "FAILING" ? "FAILING" : void 0;
-		lines.push(`[${r.classification}${flag ? ` · ${flag}` : ""}] ${r.test} (${r.file})`);
+		lines.push(`[${r.classification}${flag ? ` · ${flag}` : ""}] ${safeTriageText(r.test)} (${safeTriageText(r.file)})`);
 		for (const e of r.evidence) lines.push(`    evidence: ${e}`);
 		lines.push(`    trust:    ${r.trustVerdict}`);
 		lines.push(`    next:     ${r.nextAction}`);
@@ -2240,7 +2139,16 @@ function renderPwRunSummary(s) {
 */
 /** Hard caps (§21 threat model: parameter size + resource bounds). */
 const MAX_PARAM_BYTES = 65536;
+const MAX_DURATION_MS$1 = 36e5;
+const DEFAULT_MAX_DURATION_MS$1 = 6e5;
 const PROTOCOL_VERSION = "2025-06-18";
+const CONFIGURED_ROOT = process.env.MJOLNIR_WORKSPACE_ROOT ? resolve(process.env.MJOLNIR_WORKSPACE_ROOT) : null;
+function isPathAllowed(path) {
+	if (!CONFIGURED_ROOT) return true;
+	const candidate = resolve(path);
+	const rel = relative(CONFIGURED_ROOT, candidate);
+	return rel === "" || !rel.startsWith("..") && !isAbsolute(rel);
+}
 /** JSON-RPC error codes (subset used here). */
 const MCP_ERRORS = {
 	PARSE: -32700,
@@ -2359,7 +2267,10 @@ function validateParams(name, args) {
 	if (Buffer$1.byteLength(JSON.stringify(args), "utf8") > MAX_PARAM_BYTES) return `parameters exceed ${MAX_PARAM_BYTES} bytes (threat model §21)`;
 	if (name === "scan") {
 		if (typeof args["path"] !== "string" || args["path"].length === 0) return "scan requires a non-empty string `path`";
-		if (args["maxDurationMs"] !== void 0 && typeof args["maxDurationMs"] !== "number") return "`maxDurationMs` must be a number";
+		if (args["maxDurationMs"] !== void 0) {
+			const duration = args["maxDurationMs"];
+			if (typeof duration !== "number" || !Number.isFinite(duration) || duration <= 0 || duration > MAX_DURATION_MS$1) return "`maxDurationMs` must be a finite positive number within the server budget";
+		}
 	}
 	if (name === "explain") {
 		if (typeof args["ruleId"] !== "string" || !/^QA-[A-Z]+-\d{3}$/.test(args["ruleId"])) return "explain requires `ruleId` matching /^QA-[A-Z]+-\\d{3}$/";
@@ -2404,6 +2315,24 @@ async function handleToolCall(call) {
 			message: paramError
 		}
 	};
+	const pathArgument = call.args["path"];
+	if (typeof pathArgument === "string" && !isPathAllowed(pathArgument)) return {
+		jsonrpc: "2.0",
+		id: call.id,
+		error: {
+			code: MCP_ERRORS.INVALID_PARAMS,
+			message: "path is outside the configured workspace root"
+		}
+	};
+	const fixturesRoot = call.args["fixturesRoot"];
+	if (typeof fixturesRoot === "string" && !isPathAllowed(fixturesRoot)) return {
+		jsonrpc: "2.0",
+		id: call.id,
+		error: {
+			code: MCP_ERRORS.INVALID_PARAMS,
+			message: "fixturesRoot is outside the configured workspace root"
+		}
+	};
 	try {
 		if (call.name === "scan") {
 			const target = resolve(call.args["path"]);
@@ -2415,7 +2344,7 @@ async function handleToolCall(call) {
 					message: `scan target does not exist: ${target}`
 				}
 			};
-			const maxDurationMs = typeof call.args["maxDurationMs"] === "number" ? call.args["maxDurationMs"] : Number.POSITIVE_INFINITY;
+			const maxDurationMs = typeof call.args["maxDurationMs"] === "number" ? call.args["maxDurationMs"] : DEFAULT_MAX_DURATION_MS$1;
 			const result = await serializeScan(() => runScan$1({
 				target,
 				json: true,
@@ -3547,6 +3476,7 @@ const TRUST_LABELS = Object.fromEntries(TRUST_RUNGS.map((r) => [r.level, `${r.le
 function trustHeadline(s) {
 	if (s.level === "L5" || s.level === "L4") return s.confidence >= .75 ? "Run evidence backs these findings — trust them." : "Run evidence exists, but the scan was incomplete — verify the gaps.";
 	if (s.level === "L3") return s.confidence >= .75 ? "The relevant files executed — solid static + runtime signal." : "Files executed, evidence is thin — treat findings as leads.";
+	if (s.level === "L0") return "Static signal only, with no runtime evidence — treat as an observation, not a verdict.";
 	if (s.confidence >= .75) return "Deterministic static analysis — trustworthy, uncorroborated by a run.";
 	return "Static signal only, incomplete analysis — treat as leads, not verdicts.";
 }
@@ -3554,8 +3484,10 @@ function trustHeadline(s) {
 function trustReasons(result, s) {
 	const reasons = [];
 	const corroborated = result.findings.filter((f) => f.runtimeCorroboration !== void 0).length;
-	if (result.findings.length === 0) reasons.push(result.partial ? "No findings so far, but the scan did not finish — the gaps are not proof of cleanliness." : "No findings fired on the analyzed surface — the scan found nothing to report.");
-	else reasons.push(`${result.findings.length} finding(s) fired; ${corroborated} corroborated by a real run report.`);
+	if (result.findings.length === 0) {
+		const incomplete = result.partial || result.analysisStatus.discovery !== "complete" || result.analysisStatus.rules !== "complete" || s.level === "L0" && s.evidenceCoverage === 0;
+		reasons.push(incomplete ? "No findings so far, but the scan or evidence is incomplete — the gaps are not proof of cleanliness." : "No findings fired on the analyzed surface — the scan found nothing to report.");
+	} else reasons.push(`${result.findings.length} finding(s) fired; ${corroborated} corroborated by a real run report.`);
 	if (s.confidenceCeiling !== void 0) reasons.push(`confidence capped at ${pct$1(s.confidenceCeiling)} by: ${s.ceilingReasons.join(", ")}.`);
 	if (s.provisionalRuleIds.length > 0) reasons.push(`${s.provisionalRuleIds.length} fired rule(s) unmeasured (author-estimated FP) — disclosed, not counted.`);
 	else if (s.measuredFpOfFiredRules !== void 0) reasons.push(`every fired rule carries a measured false-positive rate (evidence-weighted ${pct$1(s.measuredFpOfFiredRules)}).`);
@@ -3580,7 +3512,7 @@ function topTrustRisks(findings, n = 3) {
 }
 /** NEXT ACTION — one concrete command, deterministic. */
 function nextAction(result) {
-	if (result.partial) return "re-run with a higher --max-duration to close the truncated surface";
+	if (result.partial || result.analysisStatus.discovery !== "complete" || result.analysisStatus.rules !== "complete" || (result.analysisStatus.rulesCrashed ?? 0) > 0) return "re-run with a higher --max-duration to close the truncated surface";
 	const risks = topTrustRisks(result.findings, 1);
 	if (risks.length === 0) return result.testDeclarationCount === 0 ? "add tests — a repo without test declarations has nothing to verify" : "no findings on the analyzed surface — install advisory CI (`mjolnir ci install`) and re-scan the changed scope";
 	return `mjolnir explain ${risks[0]?.ruleId ?? ""} — then fix the top risk first`;
@@ -3993,6 +3925,258 @@ function currentCommit(root, opts = {}) {
 	}
 }
 //#endregion
+//#region src/commands/report-io.ts
+/**
+* Shared saved-report loading (agent-handoff plan §9.0).
+*
+* `summary`, `handoff` and `why` all consume a saved `--json` report.
+* One loader, one validation, one error shape — previously
+* validateReportJson was private to summary.ts; extracting it keeps
+* the three commands byte-identical in their loading behavior without
+* duplication. The exit-code mapping stays in the callers (they own
+* their io), but the error MESSAGES are identical because they come
+* from here.
+*/
+const MAX_REPORT_BYTES = 33554432;
+const MAX_FINDINGS = 1e4;
+const CONFIDENCE_VALUES = /* @__PURE__ */ new Set([
+	"high",
+	"medium",
+	"low"
+]);
+const FINDING_TYPE_VALUES = /* @__PURE__ */ new Set([
+	"deterministic-defect",
+	"heuristic-risk",
+	"observation"
+]);
+const QA_IMPACT_VALUES = /* @__PURE__ */ new Set([
+	"BLOCKS-RELEASE",
+	"FLAKY-RISK",
+	"FALSE-GREEN",
+	"HYGIENE"
+]);
+const EVIDENCE_VALUES = /* @__PURE__ */ new Set([
+	"E0",
+	"E1",
+	"E2"
+]);
+const TRUST_VALUES = /* @__PURE__ */ new Set([
+	"L0",
+	"L1",
+	"L2",
+	"L3",
+	"L4",
+	"L5"
+]);
+/** Human message for any thrown value — never "undefined"/"[object Object]". */
+function errorText$1(err) {
+	if (err instanceof Error) return err.message;
+	if (typeof err === "string") return err;
+	if (typeof err === "object" && err !== null) return JSON.stringify(err);
+	return String(err);
+}
+function isRecord$2(value) {
+	return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+function fail(message) {
+	throw new Error(message);
+}
+function requireRecord(value, label) {
+	if (!isRecord$2(value)) fail(`${label} must be an object`);
+	return value;
+}
+function requireString(value, label) {
+	if (typeof value !== "string" || value.length === 0) fail(`${label} must be a non-empty string`);
+	return value;
+}
+function requireBoolean(value, label) {
+	if (typeof value !== "boolean") fail(`${label} must be a boolean`);
+	return value;
+}
+function requireFiniteNumber(value, label) {
+	if (typeof value !== "number" || !Number.isFinite(value)) fail(`${label} must be a finite number`);
+	return value;
+}
+function requireInteger(value, label, minimum = 0) {
+	if (typeof value !== "number" || !Number.isSafeInteger(value) || value < minimum) fail(`${label} must be an integer >= ${minimum}`);
+	return value;
+}
+function validateDimension(value, index) {
+	const dimension = requireRecord(value, `dimensions[${index}]`);
+	const category = requireString(dimension.category, `dimensions[${index}].category`);
+	if (!RULE_CATEGORIES.includes(category)) fail(`dimensions[${index}].category is unknown`);
+	const score = requireFiniteNumber(dimension.score, `dimensions[${index}].score`);
+	if (score < 0 || score > 100) fail(`dimensions[${index}].score must be between 0 and 100`);
+	requireInteger(dimension.errors, `dimensions[${index}].errors`);
+	requireInteger(dimension.warnings, `dimensions[${index}].warnings`);
+	requireInteger(dimension.infos, `dimensions[${index}].infos`);
+}
+function validateFinding(value, index) {
+	const finding = requireRecord(value, `findings[${index}]`);
+	requireString(finding.ruleId, `findings[${index}].ruleId`);
+	const category = requireString(finding.category, `findings[${index}].category`);
+	if (!RULE_CATEGORIES.includes(category)) fail(`findings[${index}].category is unknown`);
+	const severity = requireString(finding.severity, `findings[${index}].severity`);
+	if (!SEVERITY_ORDER$1.includes(severity)) fail(`findings[${index}].severity is unknown`);
+	const confidence = requireString(finding.confidence, `findings[${index}].confidence`);
+	if (!CONFIDENCE_VALUES.has(confidence)) fail(`findings[${index}].confidence is unknown`);
+	const findingType = requireString(finding.findingType, `findings[${index}].findingType`);
+	if (!FINDING_TYPE_VALUES.has(findingType)) fail(`findings[${index}].findingType is unknown`);
+	const qaImpact = requireString(finding.qaImpact, `findings[${index}].qaImpact`);
+	if (!QA_IMPACT_VALUES.has(qaImpact)) fail(`findings[${index}].qaImpact is unknown`);
+	requireString(finding.file, `findings[${index}].file`);
+	requireInteger(finding.line, `findings[${index}].line`, 1);
+	requireInteger(finding.column, `findings[${index}].column`, 1);
+	requireString(finding.message, `findings[${index}].message`);
+	requireString(finding.why, `findings[${index}].why`);
+	requireString(finding.fix, `findings[${index}].fix`);
+	if (finding.evidenceLevel !== void 0 && (typeof finding.evidenceLevel !== "string" || !EVIDENCE_VALUES.has(finding.evidenceLevel))) fail(`findings[${index}].evidenceLevel is unknown`);
+	if (finding.trustLevel !== void 0 && (typeof finding.trustLevel !== "string" || !TRUST_VALUES.has(finding.trustLevel))) fail(`findings[${index}].trustLevel is unknown`);
+	if (finding.detectorRevision !== void 0) requireInteger(finding.detectorRevision, `findings[${index}].detectorRevision`, 1);
+	if (finding.measuredFpRate !== void 0) {
+		const rate = requireFiniteNumber(finding.measuredFpRate, `findings[${index}].measuredFpRate`);
+		if (rate < 0 || rate > 1) fail(`findings[${index}].measuredFpRate must be between 0 and 1`);
+	}
+	if (finding.measuredFpN !== void 0) requireInteger(finding.measuredFpN, `findings[${index}].measuredFpN`);
+}
+function validateAnalysisStatus(value) {
+	const status = requireRecord(value, "analysisStatus");
+	if (status.discovery !== "complete" && status.discovery !== "partial") fail("analysisStatus.discovery must be \"complete\" or \"partial\"");
+	if (status.rules !== "complete" && status.rules !== "partial") fail("analysisStatus.rules must be \"complete\" or \"partial\"");
+	requireInteger(status.skippedFiles, "analysisStatus.skippedFiles");
+	requireFiniteNumber(status.durationMs, "analysisStatus.durationMs");
+	if (status.rulesCrashed !== void 0) requireInteger(status.rulesCrashed, "analysisStatus.rulesCrashed");
+	if (status.parseFallbacks !== void 0) requireInteger(status.parseFallbacks, "analysisStatus.parseFallbacks");
+	if (status.truncationReasons !== void 0) {
+		if (!Array.isArray(status.truncationReasons)) fail("analysisStatus.truncationReasons must be an array");
+		status.truncationReasons.forEach((reason, index) => requireString(reason, `analysisStatus.truncationReasons[${index}]`));
+	}
+	if (status.reasons !== void 0) {
+		if (!Array.isArray(status.reasons)) fail("analysisStatus.reasons must be an array");
+		status.reasons.forEach((reason, index) => requireString(reason, `analysisStatus.reasons[${index}]`));
+	}
+}
+function validateScopeIntegrity(value) {
+	const scope = requireRecord(value, "scopeIntegrity");
+	requireInteger(scope.discovered, "scopeIntegrity.discovered");
+	requireInteger(scope.analyzed, "scopeIntegrity.analyzed");
+	requireInteger(scope.ignored, "scopeIntegrity.ignored");
+	requireInteger(scope.unrecognized, "scopeIntegrity.unrecognized");
+	requireInteger(scope.parseFailed, "scopeIntegrity.parseFailed");
+	requireInteger(scope.truncated, "scopeIntegrity.truncated");
+	if (scope.scopeVerdict !== "PROVEN" && scope.scopeVerdict !== "PARTIAL") fail("scopeIntegrity.scopeVerdict must be \"PROVEN\" or \"PARTIAL\"");
+	if (scope.reasons !== void 0) {
+		if (!Array.isArray(scope.reasons)) fail("scopeIntegrity.reasons must be an array");
+		scope.reasons.forEach((reason, index) => requireString(reason, `scopeIntegrity.reasons[${index}]`));
+	}
+}
+/**
+* Parse-and-validate a saved report. Throws Error with a
+* human-explanatory message on invalid JSON, shape, or contract values.
+*/
+function validateReportJson(text) {
+	if (Buffer.byteLength(text, "utf8") > MAX_REPORT_BYTES) fail(`report exceeds the ${MAX_REPORT_BYTES}-byte limit`);
+	let parsed;
+	try {
+		parsed = JSON.parse(text);
+	} catch (err) {
+		throw new Error(`not valid JSON (${errorText$1(err)})`, { cause: err });
+	}
+	const doc = requireRecord(parsed, "the report is not an object");
+	if (doc.schemaVersion !== 1) fail(`unsupported schemaVersion ${JSON.stringify(doc.schemaVersion)} — expected schemaVersion 1`);
+	requireBoolean(doc.partial, "partial");
+	if (doc.score !== null) {
+		const score = requireFiniteNumber(doc.score, "score");
+		if (score < 0 || score > 100) fail("score must be null or between 0 and 100");
+	}
+	if (doc.reason !== void 0 && doc.reason !== "no-tests-found") fail("reason must be \"no-tests-found\" when present");
+	if (doc.frameworks === void 0) fail("missing a \"frameworks\" array — is this a complete Mjölnir --json report?");
+	if (!Array.isArray(doc.frameworks) || doc.frameworks.some((value) => typeof value !== "string")) fail("frameworks must be an array of strings");
+	requireBoolean(doc.frameworkDetectionUnknown, "frameworkDetectionUnknown");
+	if (!Array.isArray(doc.dimensions)) fail("dimensions must be an array");
+	doc.dimensions.forEach(validateDimension);
+	if (!Array.isArray(doc.findings)) fail("missing a \"findings\" array — is this a Mjölnir --json report?");
+	if (doc.findings.length > MAX_FINDINGS) fail(`findings exceeds the ${MAX_FINDINGS}-finding limit`);
+	doc.findings.forEach(validateFinding);
+	validateAnalysisStatus(doc.analysisStatus);
+	if (doc.scopeIntegrity !== void 0) validateScopeIntegrity(doc.scopeIntegrity);
+	const status = requireRecord(doc.analysisStatus, "analysisStatus");
+	const partialMarkers = status["discovery"] !== "complete" || status["rules"] !== "complete" || Number(status["skippedFiles"]) > 0 || Number(status["rulesCrashed"] ?? 0) > 0 || Array.isArray(status["reasons"]) && status["reasons"].length > 0;
+	if (doc.partial === false && partialMarkers) fail("partial=false conflicts with analysisStatus");
+	if (doc.partial === true && typeof doc.score === "number" && doc.score === 100) fail("partial reports cannot have a perfect score");
+	if ((isRecord$2(doc.scopeIntegrity) ? doc.scopeIntegrity : void 0)?.["scopeVerdict"] === "PARTIAL" && doc.partial === false) fail("partial=false conflicts with scopeIntegrity");
+	return parsed;
+}
+/** Load + validate a saved report from disk. Throws on any problem. */
+function loadSavedReport(reportPath) {
+	const stat = statSync(reportPath);
+	if (!stat.isFile()) fail(`report path is not a file: ${reportPath}`);
+	if (stat.size > MAX_REPORT_BYTES) fail(`report exceeds the ${MAX_REPORT_BYTES}-byte limit`);
+	return validateReportJson(readFileSync(reportPath, "utf8"));
+}
+/** True when the report file exists (callers own the not-found message). */
+function reportExists(reportPath) {
+	return existsSync(reportPath);
+}
+//#endregion
+//#region src/integrations/github/evidence-sanitization.ts
+const MAX_MARKDOWN_TEXT_LENGTH = 500;
+/**
+* Characters that have special meaning in Markdown and must be escaped
+* to render as literal text.
+*/
+const MARKDOWN_SPECIAL_RE = /[[\]*_`~\\|{}()<>#!]/g;
+/**
+* HTML tags and entities that could enable injection.
+*/
+const HTML_INJECTION_RE = /<[^>]*>/g;
+const HTML_ENTITY_RE = /&\w+;|&#\d+;|&#x[\da-f]+;/gi;
+/**
+* Patterns that could form malformed or malicious links.
+*/
+const MALFORMED_LINK_PATTERNS = [/\]\s*\(\s*(?:javascript|data|vbscript)\s*:/gi, /\]\s*\(\s*[^)]{0,200}<[^)]{0,200}>/g];
+const RTL_OVERRIDE_RE = /[\u200f\u202e\u2067\u2066\u2069\u2068]/g;
+/**
+* Escape Markdown special characters in text so they render as literals.
+*/
+function escapeMarkdownChars(text) {
+	return text.replace(MARKDOWN_SPECIAL_RE, "\\$&");
+}
+/**
+* Strip HTML tags and entities to prevent injection.
+*/
+function stripHtml(text) {
+	return text.replace(HTML_ENTITY_RE, "").replace(HTML_INJECTION_RE, "");
+}
+/**
+* Neutralize potential malformed link patterns.
+*/
+function neutralizeMalformedLinks(text) {
+	let result = text;
+	for (const pattern of MALFORMED_LINK_PATTERNS) result = result.replace(pattern, (match) => `[sanitized: ${match.slice(0, 40)}]`);
+	return result;
+}
+/**
+* Strip RTL override characters that could confuse display.
+*/
+function stripRtlOverrides(text) {
+	return text.replace(RTL_OVERRIDE_RE, "");
+}
+/**
+* Sanitize arbitrary text for safe inclusion in Markdown.
+* Escapes special chars, strips HTML, prevents link injection,
+* strips RTL overrides, and caps length.
+*/
+function sanitizeForMarkdown(text) {
+	let result = text;
+	result = stripHtml(result);
+	result = stripRtlOverrides(result);
+	result = neutralizeMalformedLinks(result);
+	result = escapeMarkdownChars(result);
+	if (result.length > MAX_MARKDOWN_TEXT_LENGTH) result = result.slice(0, MAX_MARKDOWN_TEXT_LENGTH) + "\\.\\.\\.";
+	return result;
+}
+//#endregion
 //#region src/commands/trust-report.ts
 /**
 * `mjolnir trust-report` — the Trust Artifact (Mega MVP Master Plan
@@ -4033,7 +4217,23 @@ function fallbackSummary(result) {
 }
 /** HTML-escape a text interpolation (the artifact is hostile-input-safe). */
 function esc(text) {
-	return text.replaceAll("&", "&amp;").replaceAll("<", "&lt;").replaceAll(">", "&gt;").replaceAll("\"", "&quot;");
+	return [...text].filter((character) => {
+		const code = character.codePointAt(0) ?? 0;
+		return !(code <= 31 || code >= 127 && code <= 159 || code === 8207 || code === 8238 || code >= 8294 && code <= 8297);
+	}).join("").slice(0, 1e3).replaceAll("&", "&amp;").replaceAll("<", "&lt;").replaceAll(">", "&gt;").replaceAll("\"", "&quot;");
+}
+function publicText(text) {
+	return sanitizeForMarkdown(sanitizeErrorText(text, { maxLength: 1e3 }));
+}
+function publicJsonText(text) {
+	return sanitizeErrorText(text, { maxLength: 1e3 });
+}
+function writeArtifact(path, content) {
+	try {
+		writeFileAtomic(path, content, { encoding: "utf8" });
+	} catch (err) {
+		throw new Error(`failed to write artifact ${path}: ${errorMessage(err)}`, { cause: err });
+	}
 }
 function buildArtifactIdentity(result, commit) {
 	const revisions = /* @__PURE__ */ new Map();
@@ -4061,13 +4261,13 @@ function buildArtifactIdentity(result, commit) {
 }
 /** The MD identity section lines (§9 R9). */
 function identityLinesMd(identity) {
-	const revs = identity.detectorRevisions.map((r) => `\`${r.ruleId}@${r.detectorRevision}\``).join(" · ");
+	const revs = identity.detectorRevisions.map((r) => `\`${sanitizeForMarkdown(r.ruleId)}@${r.detectorRevision}\``).join(" · ");
 	const levels = Object.entries(identity.evidenceInventory.byEvidenceLevel).map(([level, n]) => `${level}:${n}`).join(" · ");
 	return [
 		`## Artifact integrity`,
 		"",
-		`- **scanId**: ${identity.scanId ?? "unbound (pre-R9 producer — regenerate)"}`,
-		`- **Commit**: ${identity.commit ?? "unknown (not bound)"}`,
+		`- **scanId**: ${identity.scanId ? sanitizeForMarkdown(identity.scanId) : "unbound (pre-R9 producer — regenerate)"}`,
+		`- **Commit**: ${identity.commit ? sanitizeForMarkdown(identity.commit) : "unknown (not bound)"}`,
 		`- **Rule(rev) inventory**: ${revs === "" ? "none fired" : revs}`,
 		`- **Evidence inventory**: ${identity.evidenceInventory.totalFindings} finding(s), ${identity.evidenceInventory.corroborated} runtime-corroborated, by level ${levels === "" ? "none" : levels}.`,
 		""
@@ -4079,7 +4279,7 @@ function renderTrustReportMarkdown(result, label, commit) {
 	const lines = [];
 	lines.push(`<!-- mjolnir-trust-report:v1 -->`);
 	lines.push("");
-	lines.push(`# Mjölnir Trust Report — ${label}`);
+	lines.push(`# Mjölnir Trust Report — ${publicText(label)}`);
 	lines.push("");
 	lines.push(`> Tests tell you what passed. Mjölnir tells you what you can trust.`);
 	lines.push("");
@@ -4100,7 +4300,7 @@ function renderTrustReportMarkdown(result, label, commit) {
 	lines.push(`| Tests analyzed | ${result.testDeclarationCount ?? 0} in ${result.testFileCount ?? 0} files |`);
 	lines.push("");
 	if (s.ceilingReasons.length > 0) {
-		lines.push(`Incompleteness factors: ${s.ceilingReasons.join(", ")}.`);
+		lines.push(`Incompleteness factors: ${s.ceilingReasons.map(publicText).join(", ")}.`);
 		lines.push("");
 	}
 	lines.push(`## Top trust risks`);
@@ -4112,7 +4312,7 @@ function renderTrustReportMarkdown(result, label, commit) {
 		lines.push(`| ---- | -------- | -------- | ------- |`);
 		for (const f of risks) {
 			const ev = evidenceTag(f);
-			lines.push(`| ${f.ruleId} | ${f.file}:${f.line} | ${ev} | ${f.message.replaceAll("|", "\\|")} |`);
+			lines.push(`| ${publicText(f.ruleId)} | ${publicText(f.file)}:${f.line} | ${ev} | ${publicText(f.message)} |`);
 		}
 	}
 	lines.push("");
@@ -4126,7 +4326,7 @@ function renderTrustReportMarkdown(result, label, commit) {
 	lines.push("");
 	lines.push(`## Next action`);
 	lines.push("");
-	lines.push(nextAction(result));
+	lines.push(publicText(nextAction(result)));
 	lines.push("");
 	lines.push(...identityLinesMd(buildArtifactIdentity(result, commit)));
 	lines.push(`---`);
@@ -4138,6 +4338,7 @@ function renderTrustReportMarkdown(result, label, commit) {
 /** Deterministic JSON artifact (agent-consumable twin of the MD). */
 function renderTrustReportJson(result, commit) {
 	const s = fallbackSummary(result);
+	const completionVerdict = result.partial || result.analysisStatus.discovery !== "complete" || result.analysisStatus.rules !== "complete" || (result.analysisStatus.rulesCrashed ?? 0) > 0 ? "INCOMPLETE" : result.score === null ? "UNKNOWN" : "COMPLETE";
 	return JSON.stringify({
 		schemaVersion: 1,
 		artifact: "mjolnir-trust-report",
@@ -4154,6 +4355,10 @@ function renderTrustReportJson(result, commit) {
 			ceilingReasons: s.ceilingReasons
 		},
 		score: result.score,
+		partial: result.partial,
+		analysisStatus: result.analysisStatus,
+		scopeIntegrity: result.scopeIntegrity ?? null,
+		verdict: completionVerdict,
 		tests: {
 			files: result.testFileCount ?? 0,
 			declarations: result.testDeclarationCount ?? 0
@@ -4166,14 +4371,14 @@ function renderTrustReportJson(result, commit) {
 			advisory: result.findings.filter((f) => isAdvisoryFinding(f)).length
 		},
 		topTrustRisks: topTrustRisks(result.findings, 10).map((f) => ({
-			ruleId: f.ruleId,
-			file: f.file,
+			ruleId: publicJsonText(f.ruleId),
+			file: publicJsonText(f.file),
 			line: f.line,
 			severity: f.severity,
 			evidence: f.runtimeCorroboration === void 0 ? f.evidenceLevel ?? "E2" : f.runtimeCorroboration.level,
-			message: f.message
+			message: publicJsonText(f.message)
 		})),
-		nextAction: nextAction(result)
+		nextAction: publicJsonText(nextAction(result))
 	}, null, 2) + "\n";
 }
 /**
@@ -4290,26 +4495,21 @@ async function runTrustReportCommand(argv, io) {
 			io.err("error: --commit requires the run's HEAD sha");
 			return 10;
 		}
-		let raw;
-		try {
-			raw = readFileSync(resolve(fromPath), "utf8");
-		} catch (err) {
-			io.err(`error: cannot read ${fromPath}: ${errorMessage(err)}`);
-			return 10;
-		}
-		let parsed;
+		let scan;
 		try {
-			parsed = JSON.parse(raw);
+			scan = loadSavedReport(resolve(fromPath));
 		} catch (err) {
 			io.err(`error: cannot read ${fromPath}: ${errorMessage(err)}`);
 			return 10;
 		}
-		if (typeof parsed !== "object" || parsed === null || parsed.schemaVersion !== 1) {
-			io.err(`error: ${fromPath} is not a canonical mjolnir scan result (schemaVersion 1)`);
+		if (commitArg && scan.runIdentity?.commit && scan.runIdentity.commit !== commitArg) {
+			io.err(`error: report commit ${scan.runIdentity.commit} does not match --commit ${commitArg}`);
 			return 10;
 		}
-		const { contract: _contract, ...rest } = parsed;
-		const scan = rest;
+		if (commitArg && scan.runIdentity) scan.runIdentity = {
+			...scan.runIdentity,
+			commit: commitArg
+		};
 		const md = renderTrustReportMarkdown(scan, fromPath, commitArg ?? null);
 		if (argv.includes("--stdout")) {
 			io.out(md);
@@ -4319,9 +4519,9 @@ async function runTrustReportCommand(argv, io) {
 			const outPath = resolve(dirname(fromPath), TRUST_REPORT_MD);
 			const html = renderTrustReportHtml(scan, fromPath, commitArg ?? null);
 			const json = renderTrustReportJson(scan, commitArg ?? null);
-			writeFileSync(outPath, md);
-			writeFileSync(resolve(dirname(fromPath), TRUST_REPORT_HTML), html);
-			writeFileSync(resolve(dirname(fromPath), TRUST_REPORT_JSON), json);
+			writeArtifact(outPath, md);
+			writeArtifact(resolve(dirname(fromPath), TRUST_REPORT_HTML), html);
+			writeArtifact(resolve(dirname(fromPath), TRUST_REPORT_JSON), json);
 			io.out(`trust report written: ${outPath} (plus ${TRUST_REPORT_HTML}, ${TRUST_REPORT_JSON} — same identity binding)`);
 			return 0;
 		} catch (err) {
@@ -4340,17 +4540,21 @@ async function runTrustReportCommand(argv, io) {
 			target,
 			json: true,
 			verbose: false,
-			maxDurationMs: Number.POSITIVE_INFINITY,
+			maxDurationMs: 6e5,
 			scopeChanged: false,
 			format: "json"
 		});
 		const commit = currentCommit(target, { nullable: true });
+		if (commit && result.runIdentity) result.runIdentity = {
+			...result.runIdentity,
+			commit
+		};
 		const md = renderTrustReportMarkdown(result, targetArg, commit);
 		const json = renderTrustReportJson(result, commit);
 		const html = renderTrustReportHtml(result, targetArg, commit);
-		writeFileSync(join(target, TRUST_REPORT_MD), md);
-		writeFileSync(join(target, TRUST_REPORT_JSON), json);
-		writeFileSync(join(target, TRUST_REPORT_HTML), html);
+		writeArtifact(join(target, TRUST_REPORT_MD), md);
+		writeArtifact(join(target, TRUST_REPORT_JSON), json);
+		writeArtifact(join(target, TRUST_REPORT_HTML), html);
 		io.out(`trust report written: ${TRUST_REPORT_MD}, ${TRUST_REPORT_JSON}, ${TRUST_REPORT_HTML} (deterministic — same scan, same bytes)`);
 		return 0;
 	} catch (err) {
@@ -4391,10 +4595,10 @@ function findingLineMd(f) {
 	return `${icon} **${escMd(f.ruleId)}** \`${safePath}:${f.line}\` — ${escMd(f.message)} ${escMd(`[${evidenceTagLine(f)}]`)}\n  Fix: ${fixBody}`;
 }
 /** Score cell with color-coded emoji and optional delta. */
-function renderScoreBadge(score, delta) {
+function renderScoreBadge(score, delta, verdictOverride) {
 	if (score === null) return "### Score\n*n/a*";
 	deriveScoreState(score);
-	let text = `**${score}**/100 ${verdictFor(score)}`;
+	let text = `**${score}**/100 ${verdictOverride ?? verdictFor(score)}`;
 	if (delta !== void 0 && delta !== 0) text += ` (${delta > 0 ? "+" : ""}${delta})`;
 	return `### Score\n${text}`;
 }
@@ -4462,6 +4666,10 @@ function renderArtifactIntegrity(identity) {
 	lines.push("</details>");
 	return lines;
 }
+function isIncompleteResult(result) {
+	const status = result.analysisStatus;
+	return result.partial || status === void 0 || status.discovery !== "complete" || status.rules !== "complete" || (status.rulesCrashed ?? 0) > 0 || result.scopeIntegrity?.scopeVerdict === "PARTIAL";
+}
 /**
 * Render the unified "Mjölnir Verification Report" PR comment.
 *
@@ -4479,7 +4687,8 @@ function renderUnifiedReport(result, options = {}) {
 		provisionalRuleIds: [],
 		ceilingReasons: []
 	};
-	lines.push("## <img width=\"16\" src=\"https://raw.githubusercontent.com/Sergey-Bar/Mjolnir/main/assets/mjolnir-icon.svg\"/> Mjölnir Verification Report");
+	const incomplete = isIncompleteResult(result) || !result.trustSummary || summary.level === "L0" && summary.evidenceCoverage === 0;
+	lines.push("## <img width=\"16\" alt=\"Mjölnir\" src=\"https://raw.githubusercontent.com/Sergey-Bar/Mjolnir/main/assets/mjolnir-icon.svg\"/> Mjölnir Verification Report");
 	lines.push("");
 	lines.push("> Tests tell you what passed. Mjölnir tells you what you can trust.");
 	lines.push("");
@@ -4487,6 +4696,11 @@ function renderUnifiedReport(result, options = {}) {
 		lines.push("_Scope: only the lines this PR changed._");
 		lines.push("");
 	}
+	if (incomplete) {
+		const reasons = result.analysisStatus?.reasons ?? [];
+		lines.push(`> **Analysis status: INCOMPLETE.** ${reasons.length > 0 ? reasons.map((reason) => escMd(reason)).join(", ") : "The scan did not complete a trustworthy analysis."}`);
+		lines.push("");
+	}
 	const diff = options.diff;
 	const usingDiff = diff?.hasBaseline === true;
 	const findings = usingDiff ? diff.newFindings : result.findings;
@@ -4494,7 +4708,7 @@ function renderUnifiedReport(result, options = {}) {
 	const warningCount = findings.filter((f) => f.severity === "warning").length;
 	const baseScore = diff?.baselineScore;
 	const delta = baseScore !== void 0 && result.score !== null ? result.score - baseScore : void 0;
-	lines.push("<table>");
+	lines.push("<table role=\"presentation\">");
 	lines.push("<tr>");
 	lines.push("<td width=\"130\">");
 	lines.push("");
@@ -4503,7 +4717,7 @@ function renderUnifiedReport(result, options = {}) {
 	lines.push("</td>");
 	lines.push("<td width=\"130\">");
 	lines.push("");
-	lines.push(renderScoreBadge(result.score, delta));
+	lines.push(renderScoreBadge(result.score, delta, incomplete ? "INCOMPLETE" : void 0));
 	lines.push("");
 	lines.push("</td>");
 	lines.push("<td width=\"130\">");
@@ -4521,7 +4735,7 @@ function renderUnifiedReport(result, options = {}) {
 	lines.push("");
 	const state = deriveScoreState(result.score);
 	const headlineFindings = usingDiff ? findings.length : result.findings.length;
-	lines.push(`**Headline:** ${headlineFor(state, headlineFindings)}`);
+	lines.push(incomplete ? "**Headline:** Analysis is incomplete; this report is not a clean or worthy verdict." : `**Headline:** ${headlineFor(state, headlineFindings)}`);
 	lines.push("");
 	lines.push("---");
 	lines.push("");
@@ -4534,7 +4748,7 @@ function renderUnifiedReport(result, options = {}) {
 		lines.push("");
 	}
 	if (findings.length === 0) {
-		lines.push("✅ No new findings in this PR's changes.");
+		lines.push(incomplete ? "⚠️ No findings were observed, but the analysis is incomplete; this is not proof of cleanliness." : usingDiff ? "✅ No new findings in this PR's changes." : "✅ No new findings in the analyzed surface.");
 		lines.push("");
 	} else {
 		const errors = findings.filter((f) => f.severity === "error");
@@ -4599,7 +4813,7 @@ function renderUnifiedReport(result, options = {}) {
 	lines.push("```");
 	lines.push("");
 	const repoLink = options.repoUrl ? ` — ${options.repoUrl}` : "";
-	lines.push(`<sub>Posted by GitHub Actions${repoLink} — blocks merging when the CI gate reports findings at the configured severity.</sub>`);
+	lines.push(incomplete ? `<sub>Posted by GitHub Actions${repoLink} — analysis incomplete; no release verdict is claimed.</sub>` : `<sub>Posted by GitHub Actions${repoLink} — blocks merging when the CI gate reports findings at the configured severity.</sub>`);
 	return lines.join("\n");
 }
 //#endregion
@@ -4626,57 +4840,6 @@ function renderPrComment(result, options = {}) {
 	});
 }
 //#endregion
-//#region src/commands/report-io.ts
-/**
-* Shared saved-report loading (agent-handoff plan §9.0).
-*
-* `summary`, `handoff` and `why` all consume a saved `--json` report.
-* One loader, one validation, one error shape — previously
-* validateReportJson was private to summary.ts; extracting it keeps
-* the three commands byte-identical in their loading behavior without
-* duplication. The exit-code mapping stays in the callers (they own
-* their io), but the error MESSAGES are identical because they come
-* from here.
-*/
-/** Human message for any thrown value — never "undefined"/"[object Object]". */
-function errorText$1(err) {
-	if (err instanceof Error) return err.message;
-	if (typeof err === "string") return err;
-	if (typeof err === "object" && err !== null) return JSON.stringify(err);
-	return String(err);
-}
-/**
-* Parse-and-validate a saved report. Throws Error with a
-* human-explanatory message on: invalid JSON, non-object document,
-* wrong schemaVersion, missing findings array, missing frameworks
-* array (certification F1: a schema-1 report is MORE than
-* `{schemaVersion, findings}` — summary/handoff/why read
-* `result.frameworks`, so a schema-incomplete stub must be rejected
-* HERE, by the one shared loader, not crash downstream).
-*/
-function validateReportJson(text) {
-	let parsed;
-	try {
-		parsed = JSON.parse(text);
-	} catch (err) {
-		throw new Error(`not valid JSON (${errorText$1(err)})`, { cause: err });
-	}
-	if (typeof parsed !== "object" || parsed === null) throw new Error("the file is a JSON value but not an object");
-	const doc = parsed;
-	if (doc.schemaVersion !== 1) throw new Error(`unsupported schemaVersion ${JSON.stringify(doc.schemaVersion)} — expected 1`);
-	if (!Array.isArray(doc.findings)) throw new Error("missing a \"findings\" array — is this a Mjölnir --json report?");
-	if (!Array.isArray(doc.frameworks)) throw new Error("missing a \"frameworks\" array — is this a complete Mjölnir --json report?");
-	return parsed;
-}
-/** Load + validate a saved report from disk. Throws on any problem. */
-function loadSavedReport(reportPath) {
-	return validateReportJson(readFileSync(reportPath, "utf8"));
-}
-/** True when the report file exists (callers own the not-found message). */
-function reportExists(reportPath) {
-	return existsSync(reportPath);
-}
-//#endregion
 //#region src/commands/why.ts
 /**
 * `mjolnir why <file>:<line>` — occurrence-level evidence/explanation
@@ -5108,22 +5271,11 @@ function buildBadge(result, commit) {
 * (the falsifiable claim: "verified at commit X on date Y").
 */
 function renderBadgeSnippet(result, repoUrl = "https://github.com/Sergey-Bar/Mjolnir") {
-	let commit = "unknown";
-	try {
-		commit = execFileSync("git", [
-			"rev-parse",
-			"--short",
-			"HEAD"
-		], {
-			cwd: process.cwd(),
-			encoding: "utf8",
-			stdio: [
-				"ignore",
-				"pipe",
-				"ignore"
-			]
-		}).trim();
-	} catch {}
+	const commit = runGit(process.cwd(), [
+		"rev-parse",
+		"--short",
+		"HEAD"
+	])?.trim() ?? "unknown";
 	const date = (/* @__PURE__ */ new Date()).toISOString().slice(0, 10);
 	const errors = result.findings.filter((f) => f.severity === "error").length;
 	return [
@@ -5378,7 +5530,7 @@ function createRuleScaffold(input, rootDir) {
 	};
 	mkdirSync(join(rootDir, "src", "rules", parsed.dir), { recursive: true });
 	const files = [];
-	writeFileAtomic(absRule, `/**
+	const ruleSrc = `/**
  * ${input.id} — ${input.title}.
  *
  * TODO(implement): replace the placeholder below. The rule currently
@@ -5404,7 +5556,8 @@ export const ${camel(parsed.lower)} = defineRule({
     return findings;
   },
 });
-`);
+`;
+	writeFileAtomic(absRule, ruleSrc);
 	files.push(ruleRel);
 	for (const direction of ["must-fire", "must-not-fire"]) {
 		const dirRel = join("tests", "fixtures", input.id, direction);
@@ -5550,13 +5703,15 @@ function renderHandover(map) {
 * tests/privacy-network-isolation.spec.ts, which scans this file too).
 */
 const ui$13 = plainContext();
-/** The S1-resolved absolute git binary, or the bare name to fail on. */
+/** The S1-resolved absolute git binary, or null when none is available. */
 function gitExe() {
-	return resolveGitPath() ?? "git";
+	return resolveGitPath();
 }
 function git(root, args) {
+	const executable = gitExe();
+	if (!executable) return null;
 	try {
-		return execFileSync(gitExe(), [
+		return execFileSync(executable, [
 			"-C",
 			root,
 			...args
@@ -5575,8 +5730,10 @@ function git(root, args) {
 }
 /** Like git(), but returns the raw bytes — no utf8 decode round-trip. */
 function gitBuffer(root, args) {
+	const executable = gitExe();
+	if (!executable) return null;
 	try {
-		return execFileSync(gitExe(), [
+		return execFileSync(executable, [
 			"-C",
 			root,
 			...args
@@ -5610,7 +5767,16 @@ async function computeImpact(root, options) {
 		"HEAD",
 		options.baseBranch ?? "main"
 	])?.trim() ?? void 0;
-	else baseRef = git(root, ["rev-parse", baseRef])?.trim() ?? baseRef;
+	else {
+		baseRef = git(root, ["rev-parse", baseRef])?.trim();
+		if (!baseRef) return {
+			hasComparison: false,
+			unknownReason: "base-ref-invalid",
+			resolved: [],
+			introduced: [],
+			unknownFacts
+		};
+	}
 	if (!baseRef) return {
 		hasComparison: false,
 		unknownReason: "no-prior-commit",
@@ -5618,7 +5784,15 @@ async function computeImpact(root, options) {
 		introduced: [],
 		unknownFacts
 	};
-	const headRef = git(root, ["rev-parse", "HEAD"])?.trim() ?? "HEAD";
+	const headRef = git(root, ["rev-parse", "HEAD"])?.trim();
+	if (!headRef) return {
+		hasComparison: false,
+		unknownReason: "head-ref-invalid",
+		baseRef,
+		resolved: [],
+		introduced: [],
+		unknownFacts
+	};
 	if (baseRef === headRef) return {
 		hasComparison: false,
 		unknownReason: options.since ? "base-equals-head" : "no-prior-commit",
@@ -5979,7 +6153,7 @@ function tryReadPackageJson(rootDir) {
 	const p = join(rootDir, "package.json");
 	if (!existsSync(p)) return null;
 	try {
-		return parseJsonFile(readFileSync(p, "utf8"), p, isRecord$2);
+		return parseJsonFile(readFileSync(p, "utf8"), p, isRecord$3);
 	} catch {
 		return null;
 	}
@@ -7217,6 +7391,33 @@ async function runPrCommentCommand(argv, io = {
 	out,
 	err
 }) {
+	const fromIdx = argv.indexOf("--from");
+	if (fromIdx !== -1) {
+		const fromPath = argv[fromIdx + 1];
+		if (!fromPath || fromPath.startsWith("-")) {
+			io.err("error: --from requires a saved report path");
+			return 10;
+		}
+		try {
+			const saved = loadSavedReport(resolve(fromPath));
+			const option = (flag) => {
+				const index = argv.indexOf(flag);
+				return index >= 0 ? argv[index + 1] : void 0;
+			};
+			const version = option("--version");
+			const commit = option("--commit");
+			const repoUrl = option("--repo-url");
+			io.out(renderPrComment(saved, {
+				...version ? { version } : {},
+				...commit ? { commit } : {},
+				...repoUrl ? { repoUrl } : {}
+			}));
+			return 0;
+		} catch (err) {
+			io.err(`error: cannot read ${fromPath}: ${errorText$1(err)}`);
+			return 10;
+		}
+	}
 	const args = parseArgsOrUsage(argv, io);
 	if (!args) return 10;
 	try {
@@ -7376,7 +7577,12 @@ async function runRulesCommand(argv, io = {
 		return 0;
 	}
 	const root = process$1.cwd();
-	const external = withExternal ? await loadLocalRules(root) : void 0;
+	const gateOpen = pluginsGateOpen(argv.includes("--enable-plugins"));
+	const external = withExternal ? await loadLocalRules(root, gateOpen) : void 0;
+	if (external?.skipped.length) io.err(renderGateNotice(external.skipped.map((name) => ({
+		kind: "js-module",
+		name
+	}))));
 	let catalog = [...buildCatalog(), ...external ? buildCatalog(external.rules, { provenance: "external" }) : []];
 	for (const w of external?.errors ?? []) io.err(`mjolnir: ${w}`);
 	if (argv.includes("--unmeasured")) catalog = catalog.filter((e) => e.measuredFpRate === void 0);
@@ -7701,7 +7907,7 @@ function computeDetectorHashes(rules, rulesDir, retiredIds = []) {
 	return manifest;
 }
 function loadManifest(path) {
-	return parseJsonFile(readFileSync(path, "utf8"), path, (v) => isRecord$2(v));
+	return parseJsonFile(readFileSync(path, "utf8"), path, (v) => isRecord$3(v));
 }
 //#endregion
 //#region src/commands/doctor.ts
@@ -7878,7 +8084,7 @@ function checkTierEnforcement(verdictsDir, rules = RULES) {
 			const lines = readFileSync(join(verdictsDir, f), "utf8").split("\n").filter((l) => l.trim().length > 0);
 			for (const line of lines) try {
 				const parsed = JSON.parse(line);
-				if (!isRecord$2(parsed)) continue;
+				if (!isRecord$3(parsed)) continue;
 				const entry = parsed;
 				if (entry.verdict === "TP" || entry.verdict === "FP") {
 					if (typeof entry.ruleId === "string") live.set(entry.ruleId, (live.get(entry.ruleId) ?? 0) + 1);
@@ -7986,7 +8192,7 @@ function checkFixtureIntegrity(fixturesRoot, rules = RULES) {
 	const allowlistPath = join(fixturesRoot, "typecheck-allowlist.json");
 	if (existsSync(allowlistPath)) try {
 		const raw = JSON.parse(readFileSync(allowlistPath, "utf8"));
-		if (!isRecord$2(raw)) {
+		if (!isRecord$3(raw)) {
 			ok = false;
 			details.push("typecheck-allowlist.json is not an object");
 		} else if (Array.isArray(raw.entries)) details.push(`Layer A typecheck allowlist: ${raw.entries.length} justified entr(ies)`);
@@ -8100,7 +8306,7 @@ function checkMeasurementConsistency(verdictsDir, sidecarPath, rules = RULES, me
 			let row;
 			try {
 				const parsed = JSON.parse(trimmed);
-				if (!isRecord$2(parsed)) {
+				if (!isRecord$3(parsed)) {
 					failures.push(`${f}: unparseable verdict row (corpus integrity)`);
 					continue;
 				}
@@ -8117,7 +8323,7 @@ function checkMeasurementConsistency(verdictsDir, sidecarPath, rules = RULES, me
 	let sidecar = {};
 	if (existsSync(sidecarPath)) try {
 		const parsed = JSON.parse(readFileSync(sidecarPath, "utf8"));
-		if (!isRecord$2(parsed)) failures.push("sidecar is not a valid object — regenerate the measurement artifacts");
+		if (!isRecord$3(parsed)) failures.push("sidecar is not a valid object — regenerate the measurement artifacts");
 		else sidecar = parsed;
 	} catch (e) {
 		failures.push(`sidecar unreadable/malformed: ${errorText(e)} — regenerate the measurement artifacts`);
@@ -9101,19 +9307,12 @@ function hookBlock(version) {
 function resolveHookTarget(cwd) {
 	const huskyDir = join(cwd, ".husky");
 	if (existsSync(huskyDir)) return join(huskyDir, "pre-commit");
-	try {
-		const hooksPath = execFileSync("git", [
-			"-C",
-			cwd,
-			"config",
-			"core.hooksPath"
-		], { stdio: [
-			"ignore",
-			"pipe",
-			"ignore"
-		] }).toString().trim();
-		return join(cwd, hooksPath, "pre-commit");
-	} catch {}
+	const hooksPath = runGit(cwd, ["config", "core.hooksPath"])?.trim();
+	if (hooksPath) {
+		const candidate = resolve(cwd, hooksPath);
+		const relativePath = relative(cwd, candidate);
+		if (relativePath === "" || !relativePath.startsWith("..") && !isAbsolute(relativePath)) return join(candidate, "pre-commit");
+	}
 	return join(cwd, ".git", "hooks", "pre-commit");
 }
 /**
@@ -9619,7 +9818,7 @@ function checkArtifactIntegrity(root) {
 			what: "the artifact embeds its identity binding (scanId/commit/rule(rev)/evidence inventory)"
 		},
 		{
-			ok: report.includes("TRUST_REPORT_HTML") && report.includes("renderTrustReportHtml") && /writeFileSync\(join\(target,\s*TRUST_REPORT_HTML\)/.test(report),
+			ok: report.includes("TRUST_REPORT_HTML") && report.includes("renderTrustReportHtml") && /write(?:FileSync|Artifact)\(join\(target,\s*TRUST_REPORT_HTML\)/.test(report),
 			what: "the HTML artifact ships and the command writes all three formats"
 		},
 		{
@@ -10288,12 +10487,19 @@ async function runReportPlaywrightCommand(argv, io) {
 		io.err(`mjolnir report playwright: target does not exist: ${target}`);
 		return 10;
 	}
+	const root = resolve(target);
+	const requestedOutput = resolve(outputPath);
+	const outputRelative = relative(root, requestedOutput);
+	if (isAbsolute(outputRelative) || outputRelative.startsWith("..")) {
+		io.err("mjolnir report playwright: output must stay within the target root");
+		return 10;
+	}
 	try {
 		const result = await runScan$1({
 			target,
 			json: true,
 			verbose: false,
-			maxDurationMs: Number.POSITIVE_INFINITY,
+			maxDurationMs: 6e5,
 			scopeChanged: false,
 			format: "json",
 			strict: false
@@ -10301,8 +10507,8 @@ async function runReportPlaywrightCommand(argv, io) {
 		const report = buildPlaywrightReport(result);
 		const outDir = resolve(".");
 		if (!existsSync(outDir)) mkdirSync(outDir, { recursive: true });
-		const fullPath = join(outDir, outputPath);
-		writeFileSync(fullPath, JSON.stringify(report, null, 2) + "\n");
+		const fullPath = requestedOutput;
+		writeFileAtomic(fullPath, JSON.stringify(report, null, 2) + "\n", { encoding: "utf8" });
 		const header = sectionHeader("PLAYWRIGHT REPORT", ui$6);
 		io.out(`${header}\n`);
 		io.out(`Score: ${result.score !== null ? result.score + "/100" : "unknown"}`);
@@ -10787,20 +10993,44 @@ function runAnalyzeCommand(argv, io) {
 		const findings = [];
 		const fileImports = /* @__PURE__ */ new Map();
 		const fileContents = /* @__PURE__ */ new Map();
-		function walk(dir) {
+		let limited = false;
+		const deadline = Date.now() + 6e5;
+		function walk(dir, depth = 0) {
+			if (limited || depth > LIMITS.maxDepth || Date.now() >= deadline) {
+				limited = true;
+				return;
+			}
 			for (const name of readdirSync(dir)) {
 				const path = join(dir, name);
 				if (name === "node_modules" || name === ".git") continue;
-				if (statSync(path).isDirectory()) walk(path);
-				else if (name.endsWith(".ts") || name.endsWith(".tsx") || name.endsWith(".js") || name.endsWith(".jsx")) try {
-					const content = readFileSync(path, "utf8");
+				let stat;
+				try {
+					stat = lstatSync(path);
+				} catch {
+					continue;
+				}
+				if (stat.isSymbolicLink()) continue;
+				if (stat.isDirectory()) {
+					walk(path, depth + 1);
+					if (limited) return;
+				} else if (name.endsWith(".ts") || name.endsWith(".tsx") || name.endsWith(".js") || name.endsWith(".jsx")) {
+					if (fileContents.size >= LIMITS.maxFilesPerAdapter) {
+						limited = true;
+						return;
+					}
+					const read = readFileBounded(path, LIMITS.maxFileBytes);
+					if (!read.ok) {
+						limited = true;
+						continue;
+					}
+					const content = read.data.toString("utf8");
 					const rel = relative(process.cwd(), path);
 					fileContents.set(rel, content);
 					const imports = collectImports(content);
 					const importCounts = /* @__PURE__ */ new Map();
 					for (const imp of imports) importCounts.set(imp, (importCounts.get(imp) ?? 0) + 1);
 					fileImports.set(rel, importCounts);
-				} catch {}
+				}
 			}
 		}
 		walk(target);
@@ -10854,6 +11084,10 @@ function runAnalyzeCommand(argv, io) {
 			}
 			io.out("");
 		}
+		if (limited) {
+			io.err("mjolnir analyze: resource limit reached; cross-file analysis is partial");
+			return 2;
+		}
 		return 0;
 	} catch (e) {
 		internalErrorMessage(e, io.err, false);
@@ -10966,9 +11200,12 @@ function runCiAdapterCommand(argv, io) {
 * file with inline CSS and JavaScript, suitable for embedding
 * in team wikis or sharing via static hosting.
 */
+function escapeHtml(text) {
+	return sanitizeErrorText(text, { maxLength: 1e3 }).replaceAll("&", "&amp;").replaceAll("<", "&lt;").replaceAll(">", "&gt;").replaceAll("\"", "&quot;").replaceAll("'", "&#39;");
+}
 function generateDashboardHtml(data) {
 	const scoreColor = data.score === null ? "#888" : data.score >= 80 ? "#22c55e" : data.score >= 60 ? "#eab308" : "#ef4444";
-	const rows = data.findings.map((f) => `<tr><td>${f.ruleId}</td><td><span style="color:${f.severity === "error" ? "#ef4444" : f.severity === "warning" ? "#eab308" : "#22c55e"}">${f.severity}</span></td><td>${f.file}</td><td>${f.message}</td></tr>`).join("");
+	const rows = data.findings.map((f) => `<tr><td>${escapeHtml(f.ruleId)}</td><td><span style="color:${f.severity === "error" ? "#ef4444" : f.severity === "warning" ? "#eab308" : "#22c55e"}">${escapeHtml(f.severity)}</span></td><td>${escapeHtml(f.file)}</td><td>${escapeHtml(f.message)}</td></tr>`).join("");
 	return `<!DOCTYPE html>
 <html lang="en">
 <head>
@@ -10989,7 +11226,7 @@ function generateDashboardHtml(data) {
 </head>
 <body>
 <h1>🔍 Quality Dashboard</h1>
-<p>Generated: ${data.generatedAt}</p>
+<p>Generated: ${escapeHtml(data.generatedAt)}</p>
 <div class="score">${data.score !== null ? data.score + "/100" : "N/A"}</div>
 <div class="kpi-grid">
   <div class="kpi"><div class="value">${data.totalFindings}</div><div class="label">Findings</div></div>
@@ -11010,13 +11247,20 @@ async function runDashboardCommand(argv, io) {
 		io.err(`mjolnir dashboard: target does not exist: ${target}`);
 		return 10;
 	}
+	const root = resolve(target);
+	const resolvedOutput = resolve(outputPath);
+	const outputRelative = relative(root, resolvedOutput);
+	if (isAbsolute(outputRelative) || outputRelative.startsWith("..")) {
+		io.err("mjolnir dashboard: output must stay within the target root");
+		return 10;
+	}
 	try {
-		const { runScan } = await import("./scan-pipeline-C_Rs4u8N.mjs");
+		const { runScan } = await import("./scan-pipeline-CRe7-MQn.mjs");
 		const result = await runScan({
 			target,
 			json: true,
 			verbose: false,
-			maxDurationMs: Number.POSITIVE_INFINITY,
+			maxDurationMs: 6e5,
 			scopeChanged: false,
 			format: "json",
 			strict: false
@@ -11036,8 +11280,8 @@ async function runDashboardCommand(argv, io) {
 			generatedAt: (/* @__PURE__ */ new Date()).toISOString()
 		};
 		const html = generateDashboardHtml(data);
-		writeFileSync(outputPath, html);
-		io.out(`Dashboard written to ${outputPath}`);
+		writeFileAtomic(resolvedOutput, html, { encoding: "utf8" });
+		io.out(`Dashboard written to ${resolvedOutput}`);
 		io.out(`Score: ${data.score !== null ? data.score + "/100" : "N/A"}`);
 		io.out(`Findings: ${data.totalFindings} (${data.errorCount} errors, ${data.warningCount} warnings)`);
 		return data.errorCount > 0 ? 1 : 0;
@@ -14281,7 +14525,7 @@ function loadPolicy(path) {
 	}
 	let record;
 	try {
-		record = parseJsonFile(text, path, isRecord$2);
+		record = parseJsonFile(text, path, isRecord$3);
 	} catch (error) {
 		throw new ConfigValidationError(`invalid suppression policy: ${path}`, { cause: error });
 	}
@@ -14300,7 +14544,7 @@ function loadPolicy(path) {
 	};
 }
 function isTrustSnapshot(value) {
-	return isRecord$2(value) && typeof value.scanId === "string" && typeof value.timestamp === "string" && (typeof value.score === "number" || value.score === null) && typeof value.findings === "number" && typeof value.errors === "number" && typeof value.warnings === "number" && typeof value.infos === "number" && typeof value.advisory === "number" && typeof value.trustLevel === "string" && typeof value.confidence === "number" && typeof value.evidenceCoverage === "number" && typeof value.inconclusiveRate === "number" && typeof value.partial === "boolean" && typeof value.frameworkCount === "number" && typeof value.ruleCount === "number";
+	return isRecord$3(value) && typeof value.scanId === "string" && typeof value.timestamp === "string" && (typeof value.score === "number" || value.score === null) && typeof value.findings === "number" && typeof value.errors === "number" && typeof value.warnings === "number" && typeof value.infos === "number" && typeof value.advisory === "number" && typeof value.trustLevel === "string" && typeof value.confidence === "number" && typeof value.evidenceCoverage === "number" && typeof value.inconclusiveRate === "number" && typeof value.partial === "boolean" && typeof value.frameworkCount === "number" && typeof value.ruleCount === "number";
 }
 function loadTrustHistory(path) {
 	if (!existsSync(path)) return [];
@@ -14326,18 +14570,18 @@ function isStringArray(value) {
 	return Array.isArray(value) && value.every((entry) => typeof entry === "string");
 }
 function isValidAnalysisStatus(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return (value.discovery === "complete" || value.discovery === "partial") && (value.rules === "complete" || value.rules === "partial") && isNonNegativeInteger(value.skippedFiles) && isFiniteNumber(value.durationMs) && value.durationMs >= 0 && (value.truncationReasons === void 0 || isStringArray(value.truncationReasons)) && (value.rulesCrashed === void 0 || isNonNegativeInteger(value.rulesCrashed));
 }
 function isValidDimension(value) {
-	return isRecord$2(value) && typeof value.category === "string" && isFiniteNumber(value.score) && value.score >= 0 && value.score <= 100 && isNonNegativeInteger(value.errors) && isNonNegativeInteger(value.warnings) && isNonNegativeInteger(value.infos);
+	return isRecord$3(value) && typeof value.category === "string" && isFiniteNumber(value.score) && value.score >= 0 && value.score <= 100 && isNonNegativeInteger(value.errors) && isNonNegativeInteger(value.warnings) && isNonNegativeInteger(value.infos);
 }
 function isValidFinding(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return typeof value.ruleId === "string" && value.ruleId.length > 0 && typeof value.category === "string" && typeof value.file === "string" && value.file.length > 0 && isPositiveInteger(value.line) && isPositiveInteger(value.column) && typeof value.message === "string" && value.message.length > 0 && typeof value.severity === "string" && typeof value.confidence === "string" && typeof value.findingType === "string" && typeof value.qaImpact === "string" && typeof value.why === "string" && typeof value.fix === "string";
 }
 function isValidTrustSummary(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return [
 		"L0",
 		"L1",
@@ -14348,31 +14592,31 @@ function isValidTrustSummary(value) {
 	].includes(String(value.level)) && isFiniteNumber(value.confidence) && value.confidence >= 0 && value.confidence <= 1 && isFiniteNumber(value.evidenceCoverage) && value.evidenceCoverage >= 0 && value.evidenceCoverage <= 1 && isFiniteNumber(value.inconclusiveRate) && value.inconclusiveRate >= 0 && value.inconclusiveRate <= 1 && (value.measuredFpOfFiredRules === void 0 || isFiniteNumber(value.measuredFpOfFiredRules) && value.measuredFpOfFiredRules >= 0 && value.measuredFpOfFiredRules <= 1) && isStringArray(value.provisionalRuleIds) && (value.confidenceCeiling === void 0 || isFiniteNumber(value.confidenceCeiling) && value.confidenceCeiling >= 0 && value.confidenceCeiling <= 1) && isStringArray(value.ceilingReasons);
 }
 function isValidProvenance(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return isNonNegativeInteger(value.testFiles) && isNonNegativeInteger(value.generatedMarkedFiles) && isNonNegativeInteger(value.codegenLikeFiles) && isFiniteNumber(value.shareMarkedGenerated) && value.shareMarkedGenerated >= 0 && value.shareMarkedGenerated <= 1 && isNonNegativeInteger(value.findingsInGeneratedFiles) && isNonNegativeInteger(value.findingsInUnmarkedFiles) && typeof value.note === "string";
 }
 function isValidForensicVerdicts(value) {
-	if (!isRecord$2(value) || !isRecord$2(value.byVerdict)) return false;
+	if (!isRecord$3(value) || !isRecord$3(value.byVerdict)) return false;
 	return isNonNegativeInteger(value.classifications) && isNonNegativeInteger(value.inconclusive) && Object.values(value.byVerdict).every((count) => isNonNegativeInteger(count));
 }
 function isValidSummary(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return typeof value.digest === "string" && value.digest.length > 0 && isNonNegativeInteger(value.findings) && (value.score === null || isFiniteNumber(value.score) && value.score >= 0 && value.score <= 100) && isNonNegativeInteger(value.errors) && isNonNegativeInteger(value.warnings) && isNonNegativeInteger(value.infos) && isNonNegativeInteger(value.advisory);
 }
 function isValidAnnotation(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return typeof value.path === "string" && value.path.length > 0 && isPositiveInteger(value.start_line) && (value.annotation_level === "failure" || value.annotation_level === "warning" || value.annotation_level === "notice") && typeof value.message === "string" && typeof value.ruleId === "string" && value.ruleId.length > 0 && (value.detectorRevision === void 0 || isPositiveInteger(value.detectorRevision)) && typeof value.advisory === "boolean";
 }
 function isValidCompleteness(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return typeof value.partial === "boolean" && (value.discovery === "complete" || value.discovery === "partial") && (value.rules === "complete" || value.rules === "partial") && isNonNegativeInteger(value.skippedFiles) && isNonNegativeInteger(value.rulesCrashed) && isStringArray(value.truncationReasons) && typeof value.frameworkDetectionUnknown === "boolean" && isFiniteNumber(value.durationMs) && value.durationMs >= 0;
 }
 function isValidContract(value) {
-	if (!isRecord$2(value)) return false;
+	if (!isRecord$3(value)) return false;
 	return Number.isSafeInteger(value.contractVersion) && isValidSummary(value.summary) && Array.isArray(value.annotations) && value.annotations.length <= 50 && value.annotations.every((annotation) => isValidAnnotation(annotation)) && typeof value.annotationsTruncated === "boolean" && isValidCompleteness(value.completeness) && (value.trustSummary === void 0 || isValidTrustSummary(value.trustSummary)) && (value.provenance === void 0 || isValidProvenance(value.provenance)) && (value.forensicVerdicts === void 0 || isValidForensicVerdicts(value.forensicVerdicts));
 }
 function isScanResultDocument(value) {
-	if (!isRecord$2(value) || value.schemaVersion !== 1) return false;
+	if (!isRecord$3(value) || value.schemaVersion !== 1) return false;
 	return (value.score === null || isFiniteNumber(value.score) && value.score >= 0 && value.score <= 100) && typeof value.partial === "boolean" && isStringArray(value.frameworks) && Array.isArray(value.dimensions) && value.dimensions.every((dimension) => isValidDimension(dimension)) && Array.isArray(value.findings) && value.findings.every((finding) => isValidFinding(finding)) && isValidAnalysisStatus(value.analysisStatus) && (value.agenticProfile === void 0 || isValidProvenance(value.agenticProfile)) && (value.trustSummary === void 0 || isValidTrustSummary(value.trustSummary)) && (value.forensicVerdicts === void 0 || isValidForensicVerdicts(value.forensicVerdicts)) && isValidContract(value.contract);
 }
 function emitVerification(verification, json, io) {
@@ -14628,12 +14872,14 @@ async function runCIIntegrityCommand(argv, io = {
 * keeps every existing consumer stable.
 */
 const { runScan, buildUniversalRules, fallbackWorkspace, pathMatchesGlob, isValidFindingRecord, discoverAndParseRuntimeReport, KNOWN_RULE_IDS, OVERLAP_META_BY_RULE_ID, EVIDENCE_OVERRIDES, SUITE_INVALIDATING_RULE_IDS } = scan_pipeline_exports;
+const DEFAULT_MAX_DURATION_MS = 6e5;
+const MAX_DURATION_MS = 36e5;
 function parseArgs(argv, onError) {
 	const args = {
 		target: ".",
 		json: false,
 		verbose: false,
-		maxDurationMs: Number.POSITIVE_INFINITY,
+		maxDurationMs: DEFAULT_MAX_DURATION_MS,
 		scopeChanged: false,
 		format: "terminal"
 	};
@@ -14679,7 +14925,7 @@ function parseArgs(argv, onError) {
 			args.base = ref;
 		} else if (a === "--max-duration") {
 			const v = Number(argv[++i]);
-			if (!Number.isFinite(v) || v <= 0) return reject({
+			if (!Number.isFinite(v) || v <= 0 || v * 1e3 > 36e5) return reject({
 				flag: "--max-duration",
 				token: argv[i]
 			});
@@ -15047,4 +15293,4 @@ if (isEntryPoint()) try {
 	process$1.exitCode = 20;
 }
 //#endregion
-export { ENGINE_VERSION as CLI_VERSION, EVIDENCE_OVERRIDES, KNOWN_RULE_IDS, OVERLAP_META_BY_RULE_ID, SUITE_INVALIDATING_RULE_IDS, buildUniversalRules, discoverAndParseRuntimeReport, err, exitForFindings, fallbackWorkspace, internalErrorMessage, isEntryPoint, isValidFindingRecord, levenshtein, main, nearestFlags, out, parseArgs, parseArgsOrUsage, pathMatchesGlob, printUsage, renderScanOutput, runAnalyzeCommand, runBadgeCommand, runBaselineCommand, runBusinessCaseCommand, runCIIntegrityCommand, runCiAdapterCommand, runCiInstall, runContractVerifyCommand, runCreateRuleCommand, runCrossFileCommand, runDashboardCommand, runDebtCommand, runDiffCommand, runDoctorCommand, runDoctorPlaywright, runEnterpriseCommand, runEvidenceGraphCommand, runExecReportCommand, runExplainCommand, runFixCommand, runForensicsCommand, runFrameworkMaturityCommand, runHandoffCommand, runHandoverCommand, runHelpCommand, runImpactCommand, runInitCommand, runInstallCommand, runMaturityCommand, runMutationCommand, runPolicyCommand, runPrCommentCommand, runPwReportCommand, runQuarantineCommand, runReleaseReportCommand, runReleaseTrustCommand, runReportPlaywrightCommand, runRulesCommand, runScan, runScanCommand, runStatsCommand, runSummaryCommand, runSuppressionGateCommand, runSuppressions, runTrendCommand, runTriageCommand, runTrustReportCommand, runTrustTrendCommand, runVerifyCommand, runWhyCommand, usageErrorMessage, validateScanTarget };
+export { ENGINE_VERSION as CLI_VERSION, DEFAULT_MAX_DURATION_MS, EVIDENCE_OVERRIDES, KNOWN_RULE_IDS, MAX_DURATION_MS, OVERLAP_META_BY_RULE_ID, SUITE_INVALIDATING_RULE_IDS, buildUniversalRules, discoverAndParseRuntimeReport, err, exitForFindings, fallbackWorkspace, internalErrorMessage, isEntryPoint, isValidFindingRecord, levenshtein, main, nearestFlags, out, parseArgs, parseArgsOrUsage, pathMatchesGlob, printUsage, renderScanOutput, runAnalyzeCommand, runBadgeCommand, runBaselineCommand, runBusinessCaseCommand, runCIIntegrityCommand, runCiAdapterCommand, runCiInstall, runContractVerifyCommand, runCreateRuleCommand, runCrossFileCommand, runDashboardCommand, runDebtCommand, runDiffCommand, runDoctorCommand, runDoctorPlaywright, runEnterpriseCommand, runEvidenceGraphCommand, runExecReportCommand, runExplainCommand, runFixCommand, runForensicsCommand, runFrameworkMaturityCommand, runHandoffCommand, runHandoverCommand, runHelpCommand, runImpactCommand, runInitCommand, runInstallCommand, runMaturityCommand, runMutationCommand, runPolicyCommand, runPrCommentCommand, runPwReportCommand, runQuarantineCommand, runReleaseReportCommand, runReleaseTrustCommand, runReportPlaywrightCommand, runRulesCommand, runScan, runScanCommand, runStatsCommand, runSummaryCommand, runSuppressionGateCommand, runSuppressions, runTrendCommand, runTriageCommand, runTrustReportCommand, runTrustTrendCommand, runVerifyCommand, runWhyCommand, usageErrorMessage, validateScanTarget };