@gobing-ai/spur 0.4.4 → 0.4.5

package/spur.js

@@ -88794,13 +88794,12 @@ function aggregateVerifyVerdict(input) {
     return "PARTIAL";
   return "PASS";
 }
-var VERDICT_AGGREGATES, ROW_STATUSES, CHECK_SEVERITIES, CONFIDENCE_LEVELS, verdictSchema, coverageRowSchema, checkSchema, verifyVerdictSchema, NORM = (s) => String(s ?? "").toUpperCase();
+var VERDICT_AGGREGATES, ROW_STATUSES, CHECK_SEVERITIES, verdictSchema, coverageRowSchema, checkSchema, verifyVerdictSchema, NORM = (s) => String(s ?? "").toUpperCase();
 var init_verify_verdict = __esm(() => {
   init_zod();
   VERDICT_AGGREGATES = ["PASS", "PARTIAL", "FAIL", "UNKNOWN"];
   ROW_STATUSES = ["MET", "PARTIAL", "UNMET", "N/A"];
   CHECK_SEVERITIES = ["blocker", "major", "minor", "advisory"];
-  CONFIDENCE_LEVELS = ["HIGH", "MEDIUM", "LOW"];
   verdictSchema = string2().transform((v, ctx) => {
     const up = v.toUpperCase();
     if (!VERDICT_AGGREGATES.includes(up)) {
@@ -88856,18 +88855,6 @@ var init_verify_verdict = __esm(() => {
     acceptanceCriteria: array(coverageRowSchema).optional().default([]),
     checks: array(checkSchema).optional().default([]),
     source: string2().optional(),
-    confidence: string2().transform((c, ctx) => {
-      const up = c.toUpperCase();
-      if (!CONFIDENCE_LEVELS.includes(up)) {
-        ctx.addIssue({
-          code: "custom",
-          message: `invalid confidence "${c}" (expected HIGH|MEDIUM|LOW)`,
-          path: ["confidence"]
-        });
-        return "LOW";
-      }
-      return up;
-    }).optional(),
     pipelineRunId: string2().optional(),
     recordedAt: string2().optional()
   });
@@ -90130,19 +90117,6 @@ async function readVerdictFrom(fs, path, wbs) {
       };
     }
   }
-  if ("confidence" in parsed) {
-    const rawConfidence = parsed.confidence;
-    const up = typeof rawConfidence === "string" ? rawConfidence.toUpperCase() : "";
-    if (up !== "HIGH" && up !== "MEDIUM" && up !== "LOW") {
-      return {
-        artifact: undefined,
-        path,
-        readError: `invalid confidence field at ${path}: expected HIGH|MEDIUM|LOW, actual ${JSON.stringify(rawConfidence)}`,
-        missing: false
-      };
-    }
-    parsed.confidence = up;
-  }
   return { artifact: parsed, path, missing: false };
 }
 async function readVerdictArtifact(fs, runDir, wbs) {
@@ -90237,19 +90211,6 @@ function evaluateDoneTransition(input) {
   const acs = artifact.acceptanceCriteria ?? [];
   const internallyConsistentPass = artifact.verdict !== "PASS" || reqs.length > 0 || acs.length > 0;
   const effective = internallyConsistentPass ? harshnessMax(artifact.verdict, computed) : "UNKNOWN";
-  if (effective === "PASS" && artifact.confidence === undefined) {
-    return {
-      kind: "deny",
-      verdict: "PASS",
-      message: [
-        `Cannot transition task ${wbs} to done: PASS verdict does not state a confidence level.`,
-        `  task:     ${taskFilePath}`,
-        `  verdict:  ${verdictPath} (missing \`confidence\`)`,
-        `  remediation: add "confidence": "HIGH|MEDIUM|LOW" to the artifact ` + `(re-run \`/sp:dev-verify ${wbs}\`), ` + `or override with \`spur task update ${wbs} done --force-done --reason "<why>"\`.`
-      ].join(`
-`)
-    };
-  }
   if (effective === "PASS") {
     return { kind: "allow", reason: "pass" };
   }
@@ -91196,8 +91157,6 @@ function renderTesting(v) {
   lines.push("**Pipeline verify results**");
   lines.push("");
   lines.push(`- Verdict: ${v.verdict} (from verdict artifact)`);
-  if (v.confidence !== undefined)
-    lines.push(`- Confidence: ${v.confidence}`);
   lines.push("");
   if (v.requirements.length === 0) {
     lines.push("| Requirement | Status | Evidence |");
@@ -91629,11 +91588,6 @@ function lintVerifyAnswer(answerText, taskContent, featureContent) {
   } else if (!/^(PASS|PARTIAL|FAIL)$/i.test(tables.verdict.value)) {
     add(tables.verdict.line, "verdict-value", `invalid Verdict value "${tables.verdict.value}" (PASS | PARTIAL | FAIL)`);
   }
-  if (tables.confidence === null) {
-    add(0, "confidence-missing", "no `Confidence:` line (expected exactly one `Confidence: HIGH|MEDIUM|LOW` line)");
-  } else if (!/^(HIGH|MEDIUM|LOW)$/i.test(tables.confidence.value)) {
-    add(tables.confidence.line, "confidence-value", `invalid Confidence value "${tables.confidence.value}" (HIGH | MEDIUM | LOW)`);
-  }
   const reqIds = extractRequirementIds(taskContent);
   const acIndex = buildAcIdentityIndex(taskContent, featureContent ?? null);
   const seenReq = new Set;
@@ -91693,19 +91647,6 @@ function lintVerifyAnswer(answerText, taskContent, featureContent) {
         add(row.line, "evidence-hedged", hedgedMessage(`AC "${row.id.slice(0, 40)}"`, hedge));
     }
   }
-  const confidenceLevel = tables.confidence === null ? null : tables.confidence.value.toUpperCase();
-  if (confidenceLevel === "HIGH" || confidenceLevel === "LOW") {
-    const reqStatuses = tables.reqs.map((row) => normalizeReqStatus(row.status));
-    const acStatuses = tables.acs.map((row) => normalizeAcStatus(row.status));
-    const allRowsMet = reqStatuses.length > 0 && reqStatuses.every((status) => status === "MET") && acStatuses.every((status) => status === "MET");
-    const hedgedRow = [...tables.reqs, ...tables.acs].some((row) => (normalizeReqStatus(row.status) ?? normalizeAcStatus(row.status)) === "MET" && firstHedgedPhrase(row.evidence) !== null);
-    const line = tables.confidence?.line ?? 0;
-    if (confidenceLevel === "HIGH" && (!allRowsMet || hedgedRow)) {
-      add(line, "confidence-unwarranted", "Confidence HIGH is not earned by the rows: at least one row is not MET, or a MET " + "row hedges its evidence (downgrade the level, or finish the row)");
-    } else if (confidenceLevel === "LOW" && allRowsMet) {
-      add(line, "confidence-understated", "Confidence LOW understates an answer whose every row is MET (raise the level to " + "HIGH or MEDIUM)");
-    }
-  }
   return findings;
 }
 function hasClaimMarker(evidence) {
@@ -91780,7 +91721,7 @@ function normalizeEvidenceType(raw) {
   return EVIDENCE_TYPE_PRECEDENCE.find((candidate) => tokens.includes(candidate)) ?? null;
 }
 function parseAnswer(text) {
-  const out = { verdict: null, confidence: null, reqs: [], acs: [] };
+  const out = { verdict: null, reqs: [], acs: [] };
   const lines = text.split(`
 `);
   let reqTable = false;
@@ -91840,26 +91781,11 @@ function parseAnswer(text) {
     out.verdict = {
       value: verdictMatches[0]?.[1] ?? "",
       line: text.slice(0, verdictMatches[0]?.index ?? 0).split(`
-`).length
-    };
-  }
-  const confidenceMatches = [...text.matchAll(/^\s*Confidence:\s*(\S+)\s*$/gim)];
-  if (confidenceMatches.length === 1) {
-    out.confidence = {
-      value: confidenceMatches[0]?.[1] ?? "",
-      line: text.slice(0, confidenceMatches[0]?.index ?? 0).split(`
 `).length
     };
   }
   return out;
 }
-function extractAnswerConfidence(text) {
-  const matches = [...text.matchAll(/^\s*Confidence:\s*(\S+)\s*$/gim)];
-  if (matches.length !== 1)
-    return;
-  const value = (matches[0]?.[1] ?? "").toUpperCase();
-  return CONFIDENCE_LEVELS2.includes(value) ? value : undefined;
-}
 function sectionBetween(text, heading) {
   const marker = new RegExp(`^#{1,6}\\s+${heading}\\s*$`, "im");
   const match = marker.exec(text);
@@ -91969,7 +91895,7 @@ function resolveAcIdentity(rowId, index) {
   }
   return { ok: false, error: "" };
 }
-var ANSWER_LINT_MAX_FINDINGS = 10, CLAIM_MARKERS, CITATION_FORMS, HEDGED_PATTERNS, EVIDENCE_TYPE_PRECEDENCE, CONFIDENCE_LEVELS2;
+var ANSWER_LINT_MAX_FINDINGS = 10, CLAIM_MARKERS, CITATION_FORMS, HEDGED_PATTERNS, EVIDENCE_TYPE_PRECEDENCE;
 var init_verify_answer_lint = __esm(() => {
   CLAIM_MARKERS = [
     /@[a-z0-9._-]+\/[a-z0-9._-]+/i,
@@ -91993,7 +91919,6 @@ var init_verify_answer_lint = __esm(() => {
     /\bit could\b/i
   ];
   EVIDENCE_TYPE_PRECEDENCE = ["test", "command", "static-ref", "manual-review", "llm-judge", "n/a"];
-  CONFIDENCE_LEVELS2 = ["HIGH", "MEDIUM", "LOW"];
 });
 
 // ../../packages/app/src/services/feature-check.ts
@@ -114609,329 +114534,6 @@ var init_task_transition = __esm(() => {
   init_done_transition_guard();
 });
 
-// ../../packages/app/src/services/task-verdict.ts
-function confidenceCheckRow(confidence, requirements, acceptanceCriteria) {
-  const rows = [
-    ...requirements,
-    ...acceptanceCriteria ?? []
-  ];
-  const notMet = rows.filter((r) => r.status !== "MET" && r.status !== "N/A");
-  const hedged = rows.filter((r) => r.status === "MET").map((r) => firstHedgedPhrase(r.evidence ?? "")).find((phrase) => phrase !== null);
-  const level = confidence ?? "LOW";
-  const earned = confidence !== undefined && level !== "LOW" && notMet.length === 0 && hedged === undefined;
-  const basis = `${confidence === undefined ? "LOW (absent from the answer)" : level} \u2014 ${rows.length - notMet.length}/${rows.length} rows MET` + (notMet.length > 0 ? ` (not MET: ${notMet.map((r) => r.id).join(", ")})` : "") + (hedged !== undefined ? `; hedged MET evidence: "${hedged}"` : "");
-  return { name: "confidence", status: earned ? "pass" : "warn", evidence: basis };
-}
-function injectEarnedConfidenceRow(text) {
-  let parsed;
-  try {
-    parsed = JSON.parse(text);
-  } catch {
-    return text;
-  }
-  if (parsed === null || typeof parsed !== "object" || Array.isArray(parsed))
-    return text;
-  const artifact = parsed;
-  const checks = Array.isArray(artifact.checks) ? artifact.checks : [];
-  if (checks.some((c) => c !== null && typeof c === "object" && c.name === "confidence"))
-    return text;
-  const level = typeof artifact.confidence === "string" ? artifact.confidence : undefined;
-  const row = confidenceCheckRow(level, artifact.requirements ?? [], artifact.acceptanceCriteria);
-  if (row.status !== "pass")
-    return text;
-  return `${JSON.stringify({ ...artifact, checks: [...checks, row] }, null, 4)}
-`;
-}
-function deriveVerdict(answerText, taskCheckPassed) {
-  const requirements = extractRequirements(answerText);
-  const parsedAc = extractAcceptanceCriteria(answerText);
-  const acceptanceCriteria = applyAcceptanceCriteriaEvidenceRule(parsedAc.rows);
-  const checks = extractChecks(answerText, taskCheckPassed, requirements, acceptanceCriteria, parsedAc.dropped);
-  const confidence = extractAnswerConfidence(answerText);
-  const checksWithConfidence = [
-    ...checks.filter((c) => c.name !== "confidence"),
-    confidenceCheckRow(confidence, requirements, acceptanceCriteria)
-  ];
-  if (requirements.length === 0) {
-    return {
-      verdict: "UNKNOWN",
-      ...confidence !== undefined ? { confidence } : {},
-      requirements,
-      acceptanceCriteria,
-      checks: checksWithConfidence
-    };
-  }
-  const aggregate = aggregateVerifyVerdict({
-    requirements,
-    acceptanceCriteria,
-    checks: [],
-    taskCheckPassed
-  });
-  return {
-    verdict: aggregate,
-    ...confidence !== undefined ? { confidence } : {},
-    requirements,
-    acceptanceCriteria,
-    checks: checksWithConfidence
-  };
-}
-function extractRequirements(text) {
-  const reqs = [];
-  const lines = text.split(`
-`);
-  let inTable = false;
-  let colMap = null;
-  for (const line of lines) {
-    const trimmed = line.trim();
-    if (inTable && /^#{1,6}\s/.test(trimmed)) {
-      inTable = false;
-      colMap = null;
-      continue;
-    }
-    if (!trimmed.startsWith("|"))
-      continue;
-    const cells = splitTableCells2(trimmed);
-    if (!inTable && cells.length >= 2) {
-      const h0 = (cells[0] ?? "").toLowerCase().trim();
-      const h0IsId = h0.includes("req") || h0 === "requirement" || h0 === "r#" || h0 === "r" || /^r\d+$/.test(h0);
-      if (h0IsId) {
-        const statusIdx = cells.findIndex((c) => {
-          const x = c.toLowerCase().trim();
-          return x.includes("status") || x === "verdict";
-        });
-        if (statusIdx >= 0) {
-          const evidenceIdx = cells.findIndex((c) => c.toLowerCase().includes("evidence"));
-          colMap = {
-            id: 0,
-            status: statusIdx,
-            evidence: evidenceIdx >= 0 ? evidenceIdx : undefined
-          };
-          inTable = true;
-          continue;
-        }
-      }
-    }
-    if (/^[-:]+$/.test(cells[0] ?? ""))
-      continue;
-    if (inTable && colMap && cells.length >= 2) {
-      const first = (cells[0] ?? "").toLowerCase();
-      const second = (cells[1] ?? "").toLowerCase();
-      if ((first === "ac" || first.includes("acceptance") || first === "check" || first === "name") && (second.includes("status") || second === "verdict")) {
-        inTable = false;
-        colMap = null;
-        continue;
-      }
-      const id = cells[colMap.id] ?? "";
-      const statusRaw = (cells[colMap.status] ?? "").toUpperCase();
-      const evidence = colMap.evidence !== undefined ? cells[colMap.evidence] ?? "" : "";
-      const status = normalizeStatus(statusRaw);
-      if (status !== null && id.length > 0) {
-        reqs.push({ id, status, evidence });
-      }
-    }
-  }
-  return reqs;
-}
-function normalizeStatus(raw) {
-  if (/\bMET\b/.test(raw))
-    return "MET";
-  if (/\bPARTIAL\b/.test(raw))
-    return "PARTIAL";
-  if (/\bUNMET\b/.test(raw))
-    return "UNMET";
-  return null;
-}
-function splitTableCells2(row) {
-  return row.split(/(?<!\\)\|/).map((c) => c.replace(/\\\|/g, "|").trim()).filter(Boolean);
-}
-function extractAcceptanceCriteria(text) {
-  const rows = [];
-  const dropped = [];
-  const lines = text.split(`
-`);
-  let inTable = false;
-  for (const line of lines) {
-    const trimmed = line.trim();
-    if (inTable && /^#{1,6}\s/.test(trimmed)) {
-      inTable = false;
-      continue;
-    }
-    if (!trimmed.startsWith("|"))
-      continue;
-    const cells = splitTableCells2(trimmed);
-    if (!inTable && cells.length >= 4) {
-      const h0 = (cells[0] ?? "").toLowerCase();
-      const h1 = (cells[1] ?? "").toLowerCase();
-      const h2 = (cells[2] ?? "").toLowerCase();
-      if ((h0 === "ac" || h0.includes("acceptance")) && h1.includes("status") && h2.includes("evidence")) {
-        inTable = true;
-        continue;
-      }
-    }
-    if (/^[-:]+$/.test(cells[0] ?? ""))
-      continue;
-    if (inTable && cells.length >= 3) {
-      const id = cells[0] ?? "";
-      const statusRaw = cells[1] ?? "";
-      const evidenceTypeRaw = cells[2] ?? "";
-      const status = normalizeAcceptanceCriteriaStatus(statusRaw);
-      const evidenceType = normalizeEvidenceType2(evidenceTypeRaw);
-      const acShaped = cells.length >= 4 || id.length > 0 && status !== null;
-      const evidence = cells[3] ?? "";
-      if (acShaped && status !== null && evidenceType !== null && id.length > 0) {
-        rows.push({ id, status, evidenceType, evidence });
-      } else if (acShaped && id.length > 0) {
-        dropped.push(evidenceType === null && status !== null ? `${id} (unrecognised evidence type "${evidenceTypeRaw}")` : `${id} (unrecognised status "${statusRaw}")`);
-      }
-    }
-  }
-  return { rows, dropped };
-}
-function normalizeAcceptanceCriteriaStatus(raw) {
-  const upper = raw.toUpperCase();
-  if (/\bMET\b/.test(upper))
-    return "MET";
-  if (/\bPARTIAL\b/.test(upper))
-    return "PARTIAL";
-  if (/\bUNMET\b/.test(upper))
-    return "UNMET";
-  if (/\bN\/A\b/.test(upper) || /\bNA\b/.test(upper))
-    return "N/A";
-  return null;
-}
-function normalizeEvidenceTypeToken2(normalized) {
-  if (normalized === "test")
-    return "test";
-  if (normalized === "command")
-    return "command";
-  if (normalized === "static-ref" || normalized === "static" || normalized === "doc" || normalized === "docs" || normalized === "documentation") {
-    return "static-ref";
-  }
-  if (normalized === "manual-review" || normalized === "manual")
-    return "manual-review";
-  if (normalized === "llm-judge" || normalized === "judge")
-    return "llm-judge";
-  if (normalized === "n/a" || normalized === "na")
-    return "n/a";
-  return null;
-}
-function normalizeEvidenceType2(raw) {
-  const normalized = raw.toLowerCase().trim();
-  const single = normalizeEvidenceTypeToken2(normalized);
-  if (single !== null)
-    return single;
-  const tokens = normalized.split(/[+,/]/).map((part) => normalizeEvidenceTypeToken2(part.trim())).filter((token) => token !== null);
-  if (tokens.length < 2 || tokens.length !== normalized.split(/[+,/]/).filter((p) => p.trim()).length) {
-    return null;
-  }
-  return EVIDENCE_TYPE_PRECEDENCE2.find((candidate) => tokens.includes(candidate)) ?? null;
-}
-function applyAcceptanceCriteriaEvidenceRule(rows) {
-  return rows.map((row) => {
-    if (row.status === "MET" && requiresExecutableEvidence(row.id) && row.evidenceType !== "test" && row.evidenceType !== "command") {
-      return { ...row, status: "PARTIAL" };
-    }
-    return row;
-  });
-}
-function requiresExecutableEvidence(id) {
-  const normalized = id.toLowerCase();
-  const nonCore = normalized.includes("[advisory]") || normalized.includes("[non-core]");
-  const nonBehavior = normalized.includes("[non-behavior]") || normalized.includes("[docs-only]") || normalized.includes("[doc-only]");
-  return !nonCore && !nonBehavior;
-}
-function extractChecks(_text, taskCheckPassed, requirements, acceptanceCriteria, droppedAcRows = []) {
-  const checks = [
-    {
-      name: "spur task check",
-      status: taskCheckPassed ? "pass" : "fail",
-      evidence: taskCheckPassed ? "task check passed" : "task check failed"
-    }
-  ];
-  const lines = _text.split(`
-`);
-  let inChecks = false;
-  for (const line of lines) {
-    const trimmed = line.trim();
-    if (/^#{1,6}\s/.test(trimmed))
-      inChecks = false;
-    if (!trimmed.startsWith("|"))
-      continue;
-    const cells = splitTableCells2(trimmed);
-    if (cells.length >= 2) {
-      const h0 = (cells[0] ?? "").toLowerCase();
-      const h1 = (cells[1] ?? "").toLowerCase();
-      if ((h0 === "check" || h0 === "name") && h1 === "status") {
-        inChecks = true;
-        continue;
-      }
-      if (inChecks && cells.length >= 3 && !/^[-:]+$/.test(cells[0] ?? "")) {
-        checks.push({
-          name: cells[0] ?? "",
-          status: (cells[1] ?? "").toLowerCase(),
-          evidence: cells[2] ?? ""
-        });
-      }
-    }
-  }
-  if (droppedAcRows.length > 0) {
-    checks.push({
-      name: "ac-row-dropped",
-      status: "fail",
-      evidence: `${droppedAcRows.length} AC row(s) could not be parsed and were omitted from the verdict: ` + `${droppedAcRows.join("; ")}. Accepted evidence types: test, command, static-ref (aliases: ` + "static, doc, docs, documentation), manual-review, llm-judge, n/a. Accepted statuses: " + "MET, PARTIAL, UNMET, N/A."
-    });
-  }
-  const hollowMet = [
-    ...requirements.filter((r) => r.status === "MET" && r.evidence.trim() === "").map((r) => r.id),
-    ...acceptanceCriteria.filter((r) => r.status === "MET" && r.evidence.trim() === "").map((r) => r.id)
-  ];
-  if (hollowMet.length > 0) {
-    checks.push({
-      name: "hollow-met-evidence",
-      status: "fail",
-      severity: "major",
-      evidence: `${hollowMet.length} MET row(s) carry no recorded evidence and cannot support PASS: ` + `${hollowMet.join(", ")}. Re-run verify with populated evidence per row.`
-    });
-  }
-  if (acceptanceCriteria.length > 0) {
-    const downgraded = acceptanceCriteria.filter((row) => row.status === "PARTIAL" && requiresExecutableEvidence(row.id) && row.evidenceType !== "test" && row.evidenceType !== "command");
-    checks.push({
-      name: downgraded.length > 0 ? "evidence-rule-failed" : "evidence-rule-pass",
-      status: downgraded.length > 0 ? "fail" : "pass",
-      evidence: downgraded.length > 0 ? `Executable evidence missing for: ${downgraded.map((row) => row.id).join(", ")}` : "All behavior-bearing AC rows have executable evidence or are explicitly non-behavioral."
-    });
-  }
-  return checks;
-}
-function aggregateBatchVerdicts(results) {
-  const notStarted = results.filter((r) => r.outcome === "NOT-STARTED");
-  const rolledUp = results.filter((r) => r.outcome !== "NOT-STARTED");
-  const counts = {
-    PASS: rolledUp.filter((r) => r.outcome === "PASS").length,
-    PARTIAL: rolledUp.filter((r) => r.outcome === "PARTIAL").length,
-    FAIL: rolledUp.filter((r) => r.outcome === "FAIL").length,
-    UNKNOWN: rolledUp.filter((r) => r.outcome === "UNKNOWN").length
-  };
-  let verdict;
-  if (rolledUp.length === 0) {
-    verdict = "UNKNOWN";
-  } else if (counts.FAIL > 0) {
-    verdict = "FAIL";
-  } else if (counts.PARTIAL > 0 || counts.UNKNOWN > 0) {
-    verdict = "PARTIAL";
-  } else {
-    verdict = "PASS";
-  }
-  const summary = `${counts.PASS} PASS, ${counts.PARTIAL} PARTIAL, ${counts.FAIL} FAIL, ${notStarted.length} NOT-STARTED (excluded)`;
-  return { verdict, rolledUp, notStarted, summary };
-}
-var EVIDENCE_TYPE_PRECEDENCE2;
-var init_task_verdict = __esm(() => {
-  init_verify_answer_lint();
-  init_verify_verdict();
-  EVIDENCE_TYPE_PRECEDENCE2 = ["test", "command", "static-ref", "manual-review", "llm-judge", "n/a"];
-});
-
 // ../../packages/app/src/services/task-service.ts
 import { dirname as dirname32, isAbsolute as isAbsolute12, join as join57, relative as relative13, sep as sep13 } from "path";
 function citationOrigin(verdict, name) {
@@ -115680,7 +115282,7 @@ ${templateBackground}`;
       const bytes = await this.ctx.fs.readFile(verdictPath);
       ensureDurablePlaneIgnored(paths.projectRoot);
       await this.ctx.fs.ensureDir(dirname32(durablePath));
-      await atomicWriteAsync(durablePath, injectEarnedConfidenceRow(bytes), wbs, this.ctx.fs, this.ctx.projectName ?? "spur");
+      await atomicWriteAsync(durablePath, bytes, wbs, this.ctx.fs, this.ctx.projectName ?? "spur");
     }
     const result = {
       testingWritten: false,
@@ -116235,7 +115837,6 @@ var init_task_service = __esm(() => {
   init_task_section_normalizer();
   init_task_size_precheck();
   init_task_transition();
-  init_task_verdict();
   DependencyMutationError = class DependencyMutationError extends Error {
     code;
     constructor(code, message) {
@@ -116315,6 +115916,289 @@ var init_task_service = __esm(() => {
   };
 });
 
+// ../../packages/app/src/services/task-verdict.ts
+function deriveVerdict(answerText, taskCheckPassed) {
+  const requirements = extractRequirements(answerText);
+  const parsedAc = extractAcceptanceCriteria(answerText);
+  const acceptanceCriteria = applyAcceptanceCriteriaEvidenceRule(parsedAc.rows);
+  const checks = extractChecks(answerText, taskCheckPassed, requirements, acceptanceCriteria, parsedAc.dropped);
+  if (requirements.length === 0) {
+    return {
+      verdict: "UNKNOWN",
+      requirements,
+      acceptanceCriteria,
+      checks
+    };
+  }
+  const aggregate = aggregateVerifyVerdict({
+    requirements,
+    acceptanceCriteria,
+    checks: [],
+    taskCheckPassed
+  });
+  return {
+    verdict: aggregate,
+    requirements,
+    acceptanceCriteria,
+    checks
+  };
+}
+function extractRequirements(text) {
+  const reqs = [];
+  const lines = text.split(`
+`);
+  let inTable = false;
+  let colMap = null;
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (inTable && /^#{1,6}\s/.test(trimmed)) {
+      inTable = false;
+      colMap = null;
+      continue;
+    }
+    if (!trimmed.startsWith("|"))
+      continue;
+    const cells = splitTableCells2(trimmed);
+    if (!inTable && cells.length >= 2) {
+      const h0 = (cells[0] ?? "").toLowerCase().trim();
+      const h0IsId = h0.includes("req") || h0 === "requirement" || h0 === "r#" || h0 === "r" || /^r\d+$/.test(h0);
+      if (h0IsId) {
+        const statusIdx = cells.findIndex((c) => {
+          const x = c.toLowerCase().trim();
+          return x.includes("status") || x === "verdict";
+        });
+        if (statusIdx >= 0) {
+          const evidenceIdx = cells.findIndex((c) => c.toLowerCase().includes("evidence"));
+          colMap = {
+            id: 0,
+            status: statusIdx,
+            evidence: evidenceIdx >= 0 ? evidenceIdx : undefined
+          };
+          inTable = true;
+          continue;
+        }
+      }
+    }
+    if (/^[-:]+$/.test(cells[0] ?? ""))
+      continue;
+    if (inTable && colMap && cells.length >= 2) {
+      const first = (cells[0] ?? "").toLowerCase();
+      const second = (cells[1] ?? "").toLowerCase();
+      if ((first === "ac" || first.includes("acceptance") || first === "check" || first === "name") && (second.includes("status") || second === "verdict")) {
+        inTable = false;
+        colMap = null;
+        continue;
+      }
+      const id = cells[colMap.id] ?? "";
+      const statusRaw = (cells[colMap.status] ?? "").toUpperCase();
+      const evidence = colMap.evidence !== undefined ? cells[colMap.evidence] ?? "" : "";
+      const status = normalizeStatus(statusRaw);
+      if (status !== null && id.length > 0) {
+        reqs.push({ id, status, evidence });
+      }
+    }
+  }
+  return reqs;
+}
+function normalizeStatus(raw) {
+  if (/\bMET\b/.test(raw))
+    return "MET";
+  if (/\bPARTIAL\b/.test(raw))
+    return "PARTIAL";
+  if (/\bUNMET\b/.test(raw))
+    return "UNMET";
+  return null;
+}
+function splitTableCells2(row) {
+  return row.split(/(?<!\\)\|/).map((c) => c.replace(/\\\|/g, "|").trim()).filter(Boolean);
+}
+function extractAcceptanceCriteria(text) {
+  const rows = [];
+  const dropped = [];
+  const lines = text.split(`
+`);
+  let inTable = false;
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (inTable && /^#{1,6}\s/.test(trimmed)) {
+      inTable = false;
+      continue;
+    }
+    if (!trimmed.startsWith("|"))
+      continue;
+    const cells = splitTableCells2(trimmed);
+    if (!inTable && cells.length >= 4) {
+      const h0 = (cells[0] ?? "").toLowerCase();
+      const h1 = (cells[1] ?? "").toLowerCase();
+      const h2 = (cells[2] ?? "").toLowerCase();
+      if ((h0 === "ac" || h0.includes("acceptance")) && h1.includes("status") && h2.includes("evidence")) {
+        inTable = true;
+        continue;
+      }
+    }
+    if (/^[-:]+$/.test(cells[0] ?? ""))
+      continue;
+    if (inTable && cells.length >= 3) {
+      const id = cells[0] ?? "";
+      const statusRaw = cells[1] ?? "";
+      const evidenceTypeRaw = cells[2] ?? "";
+      const status = normalizeAcceptanceCriteriaStatus(statusRaw);
+      const evidenceType = normalizeEvidenceType2(evidenceTypeRaw);
+      const acShaped = cells.length >= 4 || id.length > 0 && status !== null;
+      const evidence = cells[3] ?? "";
+      if (acShaped && status !== null && evidenceType !== null && id.length > 0) {
+        rows.push({ id, status, evidenceType, evidence });
+      } else if (acShaped && id.length > 0) {
+        dropped.push(evidenceType === null && status !== null ? `${id} (unrecognised evidence type "${evidenceTypeRaw}")` : `${id} (unrecognised status "${statusRaw}")`);
+      }
+    }
+  }
+  return { rows, dropped };
+}
+function normalizeAcceptanceCriteriaStatus(raw) {
+  const upper = raw.toUpperCase();
+  if (/\bMET\b/.test(upper))
+    return "MET";
+  if (/\bPARTIAL\b/.test(upper))
+    return "PARTIAL";
+  if (/\bUNMET\b/.test(upper))
+    return "UNMET";
+  if (/\bN\/A\b/.test(upper) || /\bNA\b/.test(upper))
+    return "N/A";
+  return null;
+}
+function normalizeEvidenceTypeToken2(normalized) {
+  if (normalized === "test")
+    return "test";
+  if (normalized === "command")
+    return "command";
+  if (normalized === "static-ref" || normalized === "static" || normalized === "doc" || normalized === "docs" || normalized === "documentation") {
+    return "static-ref";
+  }
+  if (normalized === "manual-review" || normalized === "manual")
+    return "manual-review";
+  if (normalized === "llm-judge" || normalized === "judge")
+    return "llm-judge";
+  if (normalized === "n/a" || normalized === "na")
+    return "n/a";
+  return null;
+}
+function normalizeEvidenceType2(raw) {
+  const normalized = raw.toLowerCase().trim();
+  const single = normalizeEvidenceTypeToken2(normalized);
+  if (single !== null)
+    return single;
+  const tokens = normalized.split(/[+,/]/).map((part) => normalizeEvidenceTypeToken2(part.trim())).filter((token) => token !== null);
+  if (tokens.length < 2 || tokens.length !== normalized.split(/[+,/]/).filter((p) => p.trim()).length) {
+    return null;
+  }
+  return EVIDENCE_TYPE_PRECEDENCE2.find((candidate) => tokens.includes(candidate)) ?? null;
+}
+function applyAcceptanceCriteriaEvidenceRule(rows) {
+  return rows.map((row) => {
+    if (row.status === "MET" && requiresExecutableEvidence(row.id) && row.evidenceType !== "test" && row.evidenceType !== "command") {
+      return { ...row, status: "PARTIAL" };
+    }
+    return row;
+  });
+}
+function requiresExecutableEvidence(id) {
+  const normalized = id.toLowerCase();
+  const nonCore = normalized.includes("[advisory]") || normalized.includes("[non-core]");
+  const nonBehavior = normalized.includes("[non-behavior]") || normalized.includes("[docs-only]") || normalized.includes("[doc-only]");
+  return !nonCore && !nonBehavior;
+}
+function extractChecks(_text, taskCheckPassed, requirements, acceptanceCriteria, droppedAcRows = []) {
+  const checks = [
+    {
+      name: "spur task check",
+      status: taskCheckPassed ? "pass" : "fail",
+      evidence: taskCheckPassed ? "task check passed" : "task check failed"
+    }
+  ];
+  const lines = _text.split(`
+`);
+  let inChecks = false;
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (/^#{1,6}\s/.test(trimmed))
+      inChecks = false;
+    if (!trimmed.startsWith("|"))
+      continue;
+    const cells = splitTableCells2(trimmed);
+    if (cells.length >= 2) {
+      const h0 = (cells[0] ?? "").toLowerCase();
+      const h1 = (cells[1] ?? "").toLowerCase();
+      if ((h0 === "check" || h0 === "name") && h1 === "status") {
+        inChecks = true;
+        continue;
+      }
+      if (inChecks && cells.length >= 3 && !/^[-:]+$/.test(cells[0] ?? "")) {
+        checks.push({
+          name: cells[0] ?? "",
+          status: (cells[1] ?? "").toLowerCase(),
+          evidence: cells[2] ?? ""
+        });
+      }
+    }
+  }
+  if (droppedAcRows.length > 0) {
+    checks.push({
+      name: "ac-row-dropped",
+      status: "fail",
+      evidence: `${droppedAcRows.length} AC row(s) could not be parsed and were omitted from the verdict: ` + `${droppedAcRows.join("; ")}. Accepted evidence types: test, command, static-ref (aliases: ` + "static, doc, docs, documentation), manual-review, llm-judge, n/a. Accepted statuses: " + "MET, PARTIAL, UNMET, N/A."
+    });
+  }
+  const hollowMet = [
+    ...requirements.filter((r) => r.status === "MET" && r.evidence.trim() === "").map((r) => r.id),
+    ...acceptanceCriteria.filter((r) => r.status === "MET" && r.evidence.trim() === "").map((r) => r.id)
+  ];
+  if (hollowMet.length > 0) {
+    checks.push({
+      name: "hollow-met-evidence",
+      status: "fail",
+      severity: "major",
+      evidence: `${hollowMet.length} MET row(s) carry no recorded evidence and cannot support PASS: ` + `${hollowMet.join(", ")}. Re-run verify with populated evidence per row.`
+    });
+  }
+  if (acceptanceCriteria.length > 0) {
+    const downgraded = acceptanceCriteria.filter((row) => row.status === "PARTIAL" && requiresExecutableEvidence(row.id) && row.evidenceType !== "test" && row.evidenceType !== "command");
+    checks.push({
+      name: downgraded.length > 0 ? "evidence-rule-failed" : "evidence-rule-pass",
+      status: downgraded.length > 0 ? "fail" : "pass",
+      evidence: downgraded.length > 0 ? `Executable evidence missing for: ${downgraded.map((row) => row.id).join(", ")}` : "All behavior-bearing AC rows have executable evidence or are explicitly non-behavioral."
+    });
+  }
+  return checks;
+}
+function aggregateBatchVerdicts(results) {
+  const notStarted = results.filter((r) => r.outcome === "NOT-STARTED");
+  const rolledUp = results.filter((r) => r.outcome !== "NOT-STARTED");
+  const counts = {
+    PASS: rolledUp.filter((r) => r.outcome === "PASS").length,
+    PARTIAL: rolledUp.filter((r) => r.outcome === "PARTIAL").length,
+    FAIL: rolledUp.filter((r) => r.outcome === "FAIL").length,
+    UNKNOWN: rolledUp.filter((r) => r.outcome === "UNKNOWN").length
+  };
+  let verdict;
+  if (rolledUp.length === 0) {
+    verdict = "UNKNOWN";
+  } else if (counts.FAIL > 0) {
+    verdict = "FAIL";
+  } else if (counts.PARTIAL > 0 || counts.UNKNOWN > 0) {
+    verdict = "PARTIAL";
+  } else {
+    verdict = "PASS";
+  }
+  const summary = `${counts.PASS} PASS, ${counts.PARTIAL} PARTIAL, ${counts.FAIL} FAIL, ${notStarted.length} NOT-STARTED (excluded)`;
+  return { verdict, rolledUp, notStarted, summary };
+}
+var EVIDENCE_TYPE_PRECEDENCE2;
+var init_task_verdict = __esm(() => {
+  init_verify_verdict();
+  EVIDENCE_TYPE_PRECEDENCE2 = ["test", "command", "static-ref", "manual-review", "llm-judge", "n/a"];
+});
+
 // ../../packages/app/src/services/token-ledger-service.ts
 import { closeSync as closeSync6, existsSync as existsSync19, fstatSync as fstatSync2, openSync as openSync6, readSync } from "fs";
 import { join as join58 } from "path";
@@ -134429,7 +134313,7 @@ import { createRequire as createRequire3 } from "module";
 var CLI_CONFIG = {
   binaryName: "spur",
   binaryLabel: "spur",
-  binaryVersion: "0.4.4",
+  binaryVersion: "0.4.5",
   configDir: ".spur",
   configFile: ".spur/config.yaml",
   databaseFile: ".spur/spur.db"
@@ -143057,7 +142941,7 @@ await init_src3();
 import { homedir as homedir14 } from "os";
 import { basename as basename20, join as join70, relative as relative14, resolve as resolve28 } from "path";
 // ../server/package.json
-var version3 = "0.4.4";
+var version3 = "0.4.5";
 
 // ../server/src/modules/health/index.ts
 var startedAt = Date.now();