@bendyline/gezel 1.0.6 → 1.0.7

package/dist/checks/index.js

@@ -1,3 +1,33 @@
+// src/checks/workspace-exists.ts
+function createCitedPathChecker(ws) {
+  let listing = null;
+  const probes = /* @__PURE__ */ new Map();
+  const loadListing = () => {
+    listing ??= ws.list().then((files) => new Set(files.map((f) => citedPathKey(f))));
+    return listing;
+  };
+  return async (cited) => {
+    const probe = cleanCitedPath(cited);
+    if (!probe) return false;
+    if ((await loadListing()).has(probe.toLowerCase())) return true;
+    let hit = probes.get(probe);
+    if (!hit) {
+      hit = ws.read(probe).then(
+        (content) => content !== null,
+        () => false
+      );
+      probes.set(probe, hit);
+    }
+    return hit;
+  };
+}
+function cleanCitedPath(p) {
+  return p.trim().replace(/^`+|`+$/g, "").replace(/[:#].*$/, "").replace(/^\.\//, "").replace(/^\/+/, "").replace(/^workspace\//i, "");
+}
+function citedPathKey(p) {
+  return cleanCitedPath(p).toLowerCase();
+}
+
 // src/checks/html.ts
 var MIN_INLINE_JS_BYTES = 2048;
 var SCRIPT_RE = /<script\b([^>]*)>([\s\S]*?)<\/script\s*>/gi;
@@ -536,12 +566,24 @@ function valueGrounding(text, facts, opts = {}) {
     decoysDetected
   };
 }
-var DEFAULT_CITATION_RE = /\(source:\s*([^)\s]+)(?:\s+\[[^\]]*\])*\s*\)|\]\(\s*(?!#)([^)\s]+?)\s*\)|`([^`]*\/[^`]+)`/gi;
+var DEFAULT_CITATION_RE = (
+  // The inline-path form excludes newlines AND spaces on purpose: a path
+  // contains neither, and without the exclusions two failure families
+  // appear (both wild-caught). An UNBALANCED backtick lets the span swallow
+  // sentences until the next stray backtick; and even with balanced spans,
+  // the CLOSER of one legitimate span pairs with the OPENER of the next, so
+  // the prose BETWEEN two `code` spans — which mentions a path — became one
+  // giant unresolvable "citation" and honest work read as fabricated.
+  /\(source:\s*([^)\s]+)(?:\s+\[[^\]]*\])*\s*\)|\]\(\s*(?!#)([^)\s]+?)\s*\)|`([^`\s]*\/[^`\s]+)`/gi
+);
+function stripFencedBlocks(text) {
+  return text.replace(/^[ \t]*(`{3,}|~{3,})[^\n]*\n[\s\S]*?^[ \t]*\1[ \t]*$/gm, "").replace(/^[ \t]*(`{3,}|~{3,})[^\n]*\n[\s\S]*$/m, "");
+}
 function extractCitations(text, re) {
   const flags = re.flags.includes("g") ? re.flags : `${re.flags}g`;
   const global = new RegExp(re.source, flags);
   const out = [];
-  for (const m of text.matchAll(global)) {
+  for (const m of stripFencedBlocks(text).matchAll(global)) {
     const cap = m.slice(1).find((x) => x !== void 0) ?? m[0];
     if (cap) out.push(cap);
   }
@@ -550,8 +592,8 @@ function extractCitations(text, re) {
 function cleanCitation(raw) {
   return raw.trim().replace(/^[<'"`(]+/, "").replace(/[>'"`).,;:]+$/, "");
 }
-function normalizePath(p) {
-  return p.trim().toLowerCase().replace(/^\.?\//, "").replace(/^workspace\//, "");
+function normalizeForKnownMatch(p) {
+  return p.trim().replace(/\\/g, "/").replace(/^\.\//, "").replace(/^\//, "").replace(/^workspace\//i, "").replace(/\/+$/, "").toLowerCase();
 }
 async function citationsResolve(ws, file, opts = {}) {
   const content = await ws.read(file);
@@ -574,38 +616,54 @@ async function citationsResolve(ws, file, opts = {}) {
       urls: []
     };
   }
-  const cites = [...new Set(extractCitations(content, re).map(cleanCitation).filter(Boolean))];
+  const cites = [
+    ...new Set(
+      extractCitations(content, re).map(cleanCitation).filter(
+        (citation) => citation && (/^[a-z][\w+.-]*:\/\//i.test(citation) || !/[\\/]$/.test(citation))
+      )
+    )
+  ];
   const min = opts.minCitations ?? 1;
-  const listing = new Set((await ws.list()).map(normalizePath));
+  const citedPathExists = createCitedPathChecker(ws);
   const corpus = opts.corpus ? new Set(opts.corpus.map((c) => c.toLowerCase())) : null;
+  const known = new Set((opts.knownPaths ?? []).map(normalizeForKnownMatch).filter(Boolean));
   const resolved = [];
   const unresolved = [];
   const urls = [];
+  const forgiven = [];
   for (const c of cites) {
     if (/^[a-z][\w+.-]*:\/\//i.test(c) || c.startsWith("mailto:")) {
       urls.push(c);
       if (corpus && !corpus.has(c.toLowerCase())) unresolved.push(c);
       continue;
     }
-    if (listing.has(normalizePath(c)) || corpus?.has(c.toLowerCase())) resolved.push(c);
+    if (corpus?.has(c.toLowerCase()) || await citedPathExists(c)) resolved.push(c);
+    else if (known.has(normalizeForKnownMatch(c))) forgiven.push(c);
     else unresolved.push(c);
   }
-  if (cites.length < min) {
+  if (cites.length - forgiven.length < min) {
     return {
       ok: false,
-      detail: `${file} has ${cites.length} citation(s), need \u2265 ${min} \u2014 cite the source path/URL for each claim.`,
+      // Name the accepted FORMS, not just the rule: a model that wrote the
+      // right paths as plain prose ("File: src/pricing.js, line 8") reads
+      // "cite the source" as already satisfied and rewrites content instead
+      // of adding markup, looping to gate exhaustion (wild-caught:
+      // deepseek-v4 with a flawless diagnosis, three identical rejections).
+      detail: `${file} has ${cites.length - forgiven.length} recognizable citation(s), need \u2265 ${min}. Only these forms count as citations: a backticked path like \`src/file.js\`, a markdown link like [name](src/file.js), or (source: src/file.js). Plain prose paths are not counted \u2014 wrap each cited file path in backticks.`,
       resolved,
       unresolved,
-      urls
+      urls,
+      ...forgiven.length > 0 ? { forgiven } : {}
     };
   }
   if (unresolved.length > 0) {
     return {
       ok: false,
-      detail: `${file} cites ${unresolved.length} source(s) that do not exist: ${unresolved.slice(0, 5).join(", ")} \u2014 every cited path must resolve to a real file in the workspace${corpus ? "/corpus" : ""} (no fabricated citations).`,
+      detail: `${file} cites ${unresolved.length} source(s) that do not exist: ${unresolved.slice(0, 5).join(", ")}${unresolved.length > 5 ? ", \u2026" : ""} \u2014 every cited path must resolve to a real file in the workspace${corpus ? "/corpus" : ""} (no fabricated citations).`,
       resolved,
       unresolved,
-      urls
+      urls,
+      ...forgiven.length > 0 ? { forgiven } : {}
     };
   }
   return {
@@ -613,7 +671,8 @@ async function citationsResolve(ws, file, opts = {}) {
     detail: `${file} cites ${resolved.length} resolvable source(s)${urls.length ? ` (+${urls.length} URL(s) not checked offline)` : ""}`,
     resolved,
     unresolved,
-    urls
+    urls,
+    ...forgiven.length > 0 ? { forgiven } : {}
   };
 }
 function valuesSubsetOf(outputText, sourceTexts, spec) {
@@ -682,9 +741,6 @@ var VALID_SEVERITIES = /* @__PURE__ */ new Set(["critical", "high", "medium", "l
 function str(x) {
   return typeof x === "string" ? x.trim() : "";
 }
-function normalizePath2(p) {
-  return p.trim().replace(/^`+|`+$/g, "").replace(/[:#].*$/, "").replace(/^\.\//, "").replace(/^\/+/, "").replace(/^workspace\//i, "").toLowerCase();
-}
 function escapeRe(s) {
   return s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
 }
@@ -740,12 +796,12 @@ async function securityReport(ws, reportFile, opts = {}) {
   if (!arr) {
     return fail(`${findingsPath} must be a JSON array of findings (or { "findings": [...] }).`);
   }
-  const listing = new Set((await ws.list()).map(normalizePath2));
+  const citedPathExists = createCitedPathChecker(ws);
   const problems = [];
   const fabricated = [];
   let critHigh = 0;
-  arr.forEach((raw2, i) => {
-    const f = raw2 ?? {};
+  for (const [i, rawFinding] of arr.entries()) {
+    const f = rawFinding ?? {};
     const file = str(f.file ?? f.path);
     const sev = str(f.severity).toLowerCase();
     const remediation = str(f.remediation ?? f.fix ?? f.recommendation);
@@ -753,13 +809,13 @@ async function securityReport(ws, reportFile, opts = {}) {
     const line = f.line ?? f.lineStart;
     const label = file || `#${i + 1}`;
     if (!file) problems.push(`finding #${i + 1} has no file`);
-    else if (!listing.has(normalizePath2(file))) fabricated.push(file);
+    else if (!await citedPathExists(file)) fabricated.push(file);
     if (!VALID_SEVERITIES.has(sev)) problems.push(`${label} has an invalid severity "${sev}"`);
     if (!remediation) problems.push(`${label} has no remediation`);
     if (!title) problems.push(`${label} has no title/description`);
     if (typeof line !== "number") problems.push(`${label} is not pinned to a line`);
     if (sev === "critical" || sev === "high") critHigh++;
-  });
+  }
   if (fabricated.length > 0) {
     return fail(
       `findings cite ${fabricated.length} file(s) that don't exist in the workspace: ${uniq(fabricated).slice(0, 5).join(", ")} \u2014 every finding must point at a real file:line (no fabricated citations).`,
@@ -818,6 +874,191 @@ async function securityReport(ws, reportFile, opts = {}) {
   };
 }
 
+// src/checks/codebase-review-report.ts
+var DEFAULT_SECTIONS2 = [
+  "Executive summary",
+  "Scorecard",
+  "Index coverage and method",
+  "Systemic themes",
+  "Findings",
+  "Quick wins",
+  "Strategic recommendations",
+  "Verified sound",
+  "Not assessed",
+  "Suggested deeper reviews"
+];
+var VALID_SEVERITIES2 = /* @__PURE__ */ new Set(["critical", "high", "medium", "low", "info"]);
+function str2(x) {
+  return typeof x === "string" ? x.trim() : "";
+}
+function escapeRe2(s) {
+  return s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+function hasSection2(content, name) {
+  return new RegExp(`^#{1,4}\\s+${escapeRe2(name)}\\b`, "im").test(content);
+}
+function sectionBody2(content, name) {
+  const m = new RegExp(`^(#{1,4})\\s+${escapeRe2(name)}\\b`, "im").exec(content);
+  if (!m) return "";
+  const level = m[1].length;
+  const start = m.index + m[0].length;
+  const rest = content.slice(start);
+  const next = new RegExp(`^#{1,${level}}\\s+\\S`, "m").exec(rest);
+  return next ? rest.slice(0, next.index) : rest;
+}
+function countThemeItems2(section) {
+  let n = 0;
+  for (const line of section.split(/\r?\n/)) {
+    if (/^\s*(#{3,4}\s+\S|[-*]\s+\S|\d+\.\s+\S|\*\*[^*]+\*\*)/.test(line)) n++;
+  }
+  return n;
+}
+function scorecardRows(section) {
+  const lines = section.split(/\r?\n/);
+  let headerAt = -1;
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    if (/^\s*\|/.test(line) && /dimension/i.test(line) && /(grade|rating|score|health)/i.test(line)) {
+      headerAt = i;
+      break;
+    }
+  }
+  if (headerAt < 0) return -1;
+  let rows = 0;
+  for (let i = headerAt + 1; i < lines.length; i++) {
+    const line = lines[i];
+    if (!/^\s*\|/.test(line)) break;
+    if (/^\s*\|[\s:|-]+\|?\s*$/.test(line)) continue;
+    if (line.replace(/[|\s]/g, "").length === 0) continue;
+    rows++;
+  }
+  return rows;
+}
+function uniq2(xs) {
+  return [...new Set(xs)];
+}
+async function codebaseReviewReport(reports, workspace, reportFile, opts = {}) {
+  const fail = (detail, count = 0, fabricated2 = []) => ({
+    ok: false,
+    detail,
+    findingCount: count,
+    fabricated: fabricated2
+  });
+  const content = await reports.read(reportFile);
+  if (content === null) {
+    return fail(`${reportFile} not found \u2014 write the codebase review report before advancing.`);
+  }
+  const findingsPath = opts.findings ?? "codebase-review-findings.json";
+  const raw = await reports.read(findingsPath);
+  if (raw === null) {
+    return fail(
+      `${findingsPath} not found \u2014 emit a machine-readable findings JSON alongside the report.`
+    );
+  }
+  let parsed;
+  try {
+    parsed = JSON.parse(raw);
+  } catch (e) {
+    return fail(
+      `${findingsPath} is not valid JSON (${e instanceof Error ? e.message : "parse error"}). Emit a JSON array of findings.`
+    );
+  }
+  const arr = Array.isArray(parsed) ? parsed : parsed && typeof parsed === "object" && Array.isArray(parsed.findings) ? parsed.findings : null;
+  if (!arr) {
+    return fail(`${findingsPath} must be a JSON array of findings (or { "findings": [...] }).`);
+  }
+  const citedPathExists = createCitedPathChecker(workspace);
+  const problems = [];
+  const fabricated = [];
+  let critHigh = 0;
+  for (const [i, rawFinding] of arr.entries()) {
+    const f = rawFinding ?? {};
+    const file = str2(f.file ?? f.path);
+    const sev = str2(f.severity).toLowerCase();
+    const remediation = str2(f.remediation ?? f.fix ?? f.recommendation);
+    const title = str2(f.title ?? f.description ?? f.summary);
+    const line = f.line ?? f.lineStart;
+    const label = file || `#${i + 1}`;
+    if (!file) problems.push(`finding #${i + 1} has no file`);
+    else if (!await citedPathExists(file)) fabricated.push(file);
+    if (!VALID_SEVERITIES2.has(sev)) problems.push(`${label} has an invalid severity "${sev}"`);
+    if (!remediation) problems.push(`${label} has no remediation`);
+    if (!title) problems.push(`${label} has no title/description`);
+    if ((sev === "critical" || sev === "high") && typeof line !== "number") {
+      problems.push(`${label} is ${sev} but not pinned to a line`);
+    }
+    if (sev === "critical" || sev === "high") critHigh++;
+  }
+  if (fabricated.length > 0) {
+    return fail(
+      `findings cite ${fabricated.length} file(s) that don't exist in the workspace: ${uniq2(fabricated).slice(0, 5).join(", ")} \u2014 every finding must point at a real file (no fabricated citations).`,
+      arr.length,
+      uniq2(fabricated)
+    );
+  }
+  if (problems.length > 0) {
+    return fail(
+      `${problems.length} finding(s) are incomplete: ${problems.slice(0, 4).join("; ")} \u2014 every finding needs a real file, a severity, a title, and a concrete remediation (critical/high pinned to a line).`,
+      arr.length
+    );
+  }
+  const requiredSections = opts.requiredSections ?? DEFAULT_SECTIONS2;
+  const missing = requiredSections.filter((s) => !hasSection2(content, s));
+  if (missing.length > 0) {
+    return fail(
+      `the report is missing required section(s): ${missing.map((s) => `## ${s}`).join(", ")}.`,
+      arr.length
+    );
+  }
+  const minRows = opts.minScorecardRows ?? 4;
+  const rows = scorecardRows(sectionBody2(content, "Scorecard"));
+  if (rows < 0) {
+    return fail(
+      'the "Scorecard" section has no per-dimension table \u2014 add a markdown table with Dimension and Grade columns.',
+      arr.length
+    );
+  }
+  if (rows < minRows) {
+    return fail(
+      `the Scorecard table has ${rows} dimension row(s); grade at least ${minRows} dimensions (e.g. architecture, code quality, hygiene, security, tests).`,
+      arr.length
+    );
+  }
+  const themeThreshold = opts.themeThreshold ?? 3;
+  const minThemes = opts.minThemes ?? 2;
+  if (arr.length >= themeThreshold) {
+    const themes = sectionBody2(content, "Systemic themes");
+    const items = countThemeItems2(themes);
+    if (items < minThemes) {
+      return fail(
+        `the "Systemic themes" section lists ${items} theme(s); a review with ${arr.length} findings needs \u2265 ${minThemes}, each naming a root cause and its blast radius.`,
+        arr.length
+      );
+    }
+    if (!/root[\s-]?cause/i.test(themes) || !/(blast[\s-]?radius|impact|scope|reach)/i.test(themes)) {
+      return fail(
+        `the "Systemic themes" section must analyze each theme's root cause and blast radius \u2014 that analysis is absent.`,
+        arr.length
+      );
+    }
+  }
+  const summary = sectionBody2(content, "Executive summary").toLowerCase();
+  if (critHigh > 0 && /(excellent (health|shape|condition)|no (significant |material )?(issues|problems|findings)|nothing to fix|clean bill of health)/.test(
+    summary
+  ) && !/(critical|high|however|but |concern|risk)/.test(summary)) {
+    return fail(
+      `the executive summary reads clean but there ${critHigh === 1 ? "is" : "are"} ${critHigh} critical/high finding(s) \u2014 the summary must reflect the open severity.`,
+      arr.length
+    );
+  }
+  return {
+    ok: true,
+    detail: `codebase review OK: ${arr.length} finding(s), all cite real files, ${rows} scorecard dimension(s), required sections present${arr.length >= themeThreshold ? ", systemic themes analyzed" : ""}.`,
+    findingCount: arr.length,
+    fabricated: []
+  };
+}
+
 // src/checks/records.ts
 var ISO_DATE_RE = /^\d{4}-\d{2}-\d{2}$/;
 function isRealIsoDate(value) {
@@ -1803,14 +2044,184 @@ function findCycle(rows) {
   }
   return cycle;
 }
+
+// src/checks/corpus-coverage.ts
+function failure(detail) {
+  return {
+    ok: false,
+    detail,
+    expectedBatches: 0,
+    mergedBatches: 0,
+    missingBatches: []
+  };
+}
+function parseStringArray(value) {
+  if (!Array.isArray(value) || value.some((item) => typeof item !== "string" || item === "")) {
+    return null;
+  }
+  return value;
+}
+function firstMismatch(actual, expected) {
+  if (actual.length !== expected.length) return Math.min(actual.length, expected.length);
+  for (let index = 0; index < expected.length; index += 1) {
+    if (actual[index] !== expected[index]) return index;
+  }
+  return null;
+}
+function mergeCorpusCoverageShards(batchesContent, shards, opts = {}) {
+  const batchesFile = opts.batchesFile ?? "batches.json";
+  let batchesRaw;
+  try {
+    batchesRaw = JSON.parse(batchesContent);
+  } catch (err) {
+    return failure(
+      `${batchesFile} is not valid JSON (${err instanceof Error ? err.message : String(err)}).`
+    );
+  }
+  if (!Array.isArray(batchesRaw) || batchesRaw.length === 0) {
+    return failure(`${batchesFile} must contain a non-empty batch array.`);
+  }
+  const batches = [];
+  const allPaths = /* @__PURE__ */ new Set();
+  const allRecords = /* @__PURE__ */ new Set();
+  for (let index = 0; index < batchesRaw.length; index += 1) {
+    const raw = batchesRaw[index];
+    if (typeof raw !== "object" || raw === null || Array.isArray(raw)) {
+      return failure(`${batchesFile}[${index}] must be an object.`);
+    }
+    const fields = raw;
+    const batchNumber = fields.batchNumber;
+    const paths = parseStringArray(fields.paths);
+    const records = parseStringArray(fields.records);
+    if (!Number.isInteger(batchNumber) || batchNumber !== index + 1) {
+      return failure(
+        `${batchesFile}[${index}] must have batchNumber ${index + 1}; found ${JSON.stringify(batchNumber)}.`
+      );
+    }
+    if (!paths || paths.length === 0) {
+      return failure(`${batchesFile}[${index}].paths must be a non-empty string array.`);
+    }
+    if (!records || records.length !== paths.length) {
+      return failure(
+        `${batchesFile}[${index}].records must contain one exact record path per changed path.`
+      );
+    }
+    for (const path of paths) {
+      if (allPaths.has(path))
+        return failure(`${batchesFile} assigns '${path}' to multiple batches.`);
+      allPaths.add(path);
+    }
+    for (const record of records) {
+      if (allRecords.has(record)) {
+        return failure(`${batchesFile} assigns record '${record}' to multiple batches.`);
+      }
+      allRecords.add(record);
+    }
+    batches.push({ batchNumber, paths, records });
+  }
+  const shardByBatch = /* @__PURE__ */ new Map();
+  for (const shard of shards) {
+    const normalized = shard.path.replace(/\\/g, "/");
+    const match = /(?:^|\/)coverage-(\d+)\.json$/.exec(normalized);
+    if (!match) continue;
+    const batchNumber = Number(match[1]);
+    if (batchNumber < 1 || batchNumber > batches.length) {
+      return failure(
+        `${shard.path} claims batch ${batchNumber}, but ${batchesFile} has batches 1-${batches.length}.`
+      );
+    }
+    if (shardByBatch.has(batchNumber)) {
+      return failure(`More than one coverage shard claims batch ${batchNumber}.`);
+    }
+    shardByBatch.set(batchNumber, shard);
+  }
+  const reviewedFiles = [];
+  const reviewedRecords = [];
+  const sources = [];
+  for (const batch of batches) {
+    const shard = shardByBatch.get(batch.batchNumber);
+    if (!shard) continue;
+    let parsed;
+    try {
+      parsed = JSON.parse(shard.content);
+    } catch (err) {
+      return failure(
+        `${shard.path} is not valid JSON (${err instanceof Error ? err.message : String(err)}).`
+      );
+    }
+    if (typeof parsed !== "object" || parsed === null || Array.isArray(parsed)) {
+      return failure(`${shard.path} must contain a JSON object.`);
+    }
+    const fields = parsed;
+    if (fields.batchNumber !== batch.batchNumber) {
+      return failure(
+        `${shard.path} must declare batchNumber ${batch.batchNumber}; found ${JSON.stringify(fields.batchNumber)}.`
+      );
+    }
+    const files = parseStringArray(fields.reviewedFiles);
+    const records = parseStringArray(fields.reviewedRecords);
+    if (!files || !records) {
+      return failure(`${shard.path} must contain reviewedFiles and reviewedRecords string arrays.`);
+    }
+    const fileMismatch = firstMismatch(files, batch.paths);
+    if (fileMismatch !== null) {
+      return failure(
+        `${shard.path} does not exactly match batch ${batch.batchNumber}'s changed paths at position ${fileMismatch + 1}; coverage may only come from that batch's shard.`
+      );
+    }
+    const recordMismatch = firstMismatch(records, batch.records);
+    if (recordMismatch !== null) {
+      return failure(
+        `${shard.path} does not exactly match batch ${batch.batchNumber}'s artifact records at position ${recordMismatch + 1}.`
+      );
+    }
+    reviewedFiles.push(...files);
+    reviewedRecords.push(...records);
+    sources.push({
+      batchNumber: batch.batchNumber,
+      shard: shard.path.replace(/\\/g, "/")
+    });
+  }
+  const missingBatches = batches.map((batch) => batch.batchNumber).filter((batchNumber) => !shardByBatch.has(batchNumber));
+  if (opts.requireComplete && missingBatches.length > 0) {
+    return {
+      ok: false,
+      detail: `Coverage shards are still missing for batch${missingBatches.length === 1 ? "" : "es"} ${missingBatches.join(", ")}.`,
+      expectedBatches: batches.length,
+      mergedBatches: sources.length,
+      missingBatches
+    };
+  }
+  const ledger = {
+    ...opts.pullRequest !== void 0 ? { pullRequest: opts.pullRequest } : {},
+    reviewedFiles,
+    reviewedRecords,
+    sources,
+    complete: missingBatches.length === 0
+  };
+  return {
+    ok: true,
+    detail: missingBatches.length === 0 ? `Coverage is provenance-complete across all ${batches.length} batches and ${reviewedFiles.length} changed paths.` : `Merged ${sources.length}/${batches.length} coverage shards; waiting for batches ${missingBatches.join(", ")}.`,
+    ledger,
+    content: `${JSON.stringify(ledger, null, 2)}
+`,
+    expectedBatches: batches.length,
+    mergedBatches: sources.length,
+    missingBatches
+  };
+}
 export {
   IMG_EXT,
   MIN_INLINE_JS_BYTES,
   MIN_JUDGE_EVIDENCE_SUBSTRING,
   buildJudgePrompt,
   citationsResolve,
+  citedPathKey,
+  cleanCitedPath,
+  codebaseReviewReport,
   containsPattern,
   countDistinctMatches,
+  createCitedPathChecker,
   cssMinBytes,
   csvShape,
   dataTableSniff,
@@ -1832,6 +2243,7 @@ export {
   jsonPathEquals,
   jsonValid,
   markdownHeadingsMatch,
+  mergeCorpusCoverageShards,
   namedEntitiesConsistent,
   normalizeDigitGroups,
   notContainsPattern,