gsd-pi 2.80.0-dev.c5f2443b3 → 2.80.0-dev.f55d16d13

package/src/resources/extensions/gsd/pr-evidence.ts

@@ -11,6 +11,7 @@ export interface PrEvidenceInput {
   changeType?: PrChangeType;
   linkedIssue?: string;
   summaries?: string[];
+  blockers?: string[];
   roadmapItems?: string[];
   metrics?: string[];
   testsRun?: string[];
@@ -34,8 +35,54 @@ const CHANGE_TYPE_LABELS: Record<PrChangeType, string> = {
   chore: "Build, CI, or tooling changes",
 };
 
+// Per-item cap for user-supplied content. 2 KB gives slice titles plus
+// descriptions room while still bounding malicious DoS-via-PR-body input.
+const USER_CONTENT_CAP_BYTES = 2048;
+const TRUNCATION_SUFFIX = " … [truncated]";
+
+// Strips HTML comments, fake commit trailers (Co-Authored-By, Signed-off-by —
+// case-insensitive on the trailer name), and caps total length. Designed to
+// be a no-op for well-formed input; golden fixtures must remain byte-stable.
+// Trailer lines are removed (not rejected) so that a single bad line in an
+// otherwise legitimate description does not block the entire PR.
+function sanitizeUserContent(s: string): string {
+  if (!s) return s;
+  // Strip HTML comments greedily across newlines.
+  let out = s.replace(/<!--[\s\S]*?-->/g, "");
+  // Drop lines that look like commit trailers we do not want forged.
+  out = out
+    .split("\n")
+    .filter((line) => !/^\s*(co-authored-by|signed-off-by)\s*:/i.test(line))
+    .join("\n");
+  if (Buffer.byteLength(out, "utf8") > USER_CONTENT_CAP_BYTES) {
+    const budget = USER_CONTENT_CAP_BYTES - Buffer.byteLength(TRUNCATION_SUFFIX, "utf8");
+    // Truncate by code units to stay safely under the byte budget for ASCII;
+    // for multibyte content we conservatively walk back until under budget.
+    let sliced = out.slice(0, Math.max(0, budget));
+    while (Buffer.byteLength(sliced, "utf8") > budget && sliced.length > 0) {
+      sliced = sliced.slice(0, -1);
+    }
+    out = sliced + TRUNCATION_SUFFIX;
+  }
+  return out;
+}
+
+// Strips HTML comments and fake trailers without applying the length cap.
+// Used for short fields like linkedIssue where truncation would be confusing.
+function sanitizeIssueRef(s: string): string {
+  if (!s) return s;
+  let out = s.replace(/<!--[\s\S]*?-->/g, "");
+  out = out
+    .split("\n")
+    .filter((line) => !/^\s*(co-authored-by|signed-off-by)\s*:/i.test(line))
+    .join("\n");
+  return out;
+}
+
 function normalizeList(values: readonly string[] | undefined): string[] {
-  return (values ?? []).map((value) => value.trim()).filter(Boolean);
+  return (values ?? [])
+    .map((value) => sanitizeUserContent(value).trim())
+    .filter(Boolean);
 }
 
 function changeTypeChecklist(selected: PrChangeType): string[] {
@@ -56,13 +103,17 @@ export function buildPrEvidence(input: PrEvidenceInput): PrEvidence {
   const subjectTitle = input.milestoneTitle?.trim() || subjectId;
   const changeType = input.changeType ?? "feat";
   const summaries = normalizeList(input.summaries);
+  const blockers = normalizeList(input.blockers);
   const roadmapItems = normalizeList(input.roadmapItems);
   const metrics = normalizeList(input.metrics);
   const testsRun = normalizeList(input.testsRun);
   const rollbackNotes = normalizeList(input.rollbackNotes);
-  const linkedIssue = input.linkedIssue?.trim() || "Not specified. Add an issue link before marking this PR ready if CONTRIBUTING.md requires one.";
-  const why = input.why?.trim() || `${capitalize(subjectKind)} work is complete and ready for review.`;
-  const how = input.how?.trim() || "Generated from GSD evidence and local workflow artifacts.";
+  // linkedIssue is sanitized but not length-capped: legitimate issue refs
+  // are short by nature, and truncating "Closes #123" would be unhelpful.
+  const linkedIssueRaw = input.linkedIssue ? sanitizeIssueRef(input.linkedIssue).trim() : "";
+  const linkedIssue = linkedIssueRaw || "Not specified. Add an issue link before marking this PR ready if CONTRIBUTING.md requires one.";
+  const why = (input.why ? sanitizeUserContent(input.why).trim() : "") || `${capitalize(subjectKind)} work is complete and ready for review.`;
+  const how = (input.how ? sanitizeUserContent(input.how).trim() : "") || "Generated from GSD evidence and local workflow artifacts.";
   const title = `${changeType}: ${subjectTitle}`;
 
   const sections: string[] = [
@@ -75,6 +126,13 @@ export function buildPrEvidence(input: PrEvidenceInput): PrEvidence {
     "## What",
     "",
     summaries.length > 0 ? summaries.join("\n\n") : `${capitalize(subjectKind)} ${subjectId} completed.`,
+  ];
+
+  if (blockers.length > 0) {
+    sections.push("", "## Blockers", "", blockers.map((blocker) => `- ${blocker}`).join("\n"));
+  }
+
+  sections.push(
     "",
     "## Why",
     "",
@@ -87,7 +145,7 @@ export function buildPrEvidence(input: PrEvidenceInput): PrEvidence {
     "## Linked Issue",
     "",
     linkedIssue,
-  ];
+  );
 
   if (roadmapItems.length > 0) {
     sections.push("", "## Roadmap", "", roadmapItems.join("\n"));

package/src/resources/extensions/gsd/prompts/complete-milestone.md

@@ -16,15 +16,14 @@ Start with what the excerpts give you. Read full files when the section heads si
 
 **On-demand Read ordering:** Complete all slice SUMMARY Reads you need for cross-slice synthesis, the Decision Re-evaluation table, and LEARNINGS **before** calling `gsd_complete_milestone` (step 12). Once that tool runs, the milestone is marked complete in the DB, so it must be the final persistent milestone-closeout write.
 
-### Delegate Review Work
+### Closeout Review Mode
 
-Use `subagent` for review work needing fresh context, before drafting LEARNINGS:
+The inlined context includes a validation status block.
 
-- Cross-slice integrations or new public APIs -> **reviewer** with milestone diff and roadmap.
-- Auth, network, parsing, file IO, shell exec, or crypto -> **security** audit.
-- Significant tests added or changed -> **tester** coverage check against success criteria.
+- If it says a passing validation artifact is present, treat that artifact as authoritative for success criteria, requirement coverage, verification classes, and cross-slice integration. Do not delegate fresh reviewer/security/tester audits unless the validation artifact is internally inconsistent with the inlined summaries.
+- If validation is missing, stale, non-pass, or internally inconsistent, use `subagent` for review work needing fresh context before drafting LEARNINGS: cross-slice integrations or new public APIs -> **reviewer**; auth, network, parsing, file IO, shell exec, or crypto -> **security**; significant tests added or changed -> **tester**.
 
-Subagents report only; they do not write user source. Fold findings into Decision Re-evaluation and LEARNINGS before completion.
+Subagents report only; they do not write user source. Fold any findings into Decision Re-evaluation and LEARNINGS before completion.
 
 {{inlinedContext}}
 
@@ -33,8 +32,8 @@ Subagents report only; they do not write user source. Fold findings into Decisio
 1. Use the **Milestone Summary** output template from the inlined context above
 2. {{skillActivation}}
 3. **Verify code changes exist.** Compare milestone work against the integration branch (`main`, `master`, or recorded branch), using merge-base as older revision and `HEAD` as newer. If the diff lists non-`.gsd/` files, pass. If `HEAD` equals the integration branch/merge-base, treat it as a self-diff retry: inspect milestone-scoped commit evidence (`GSD-Unit: {{milestoneId}}` or production `GSD-Task: Sxx/Tyy` trailers touching `.gsd/milestones/{{milestoneId}}/`) and verify those commits touched non-`.gsd/` files. Record **verification failure** only when neither source shows implementation files.
-4. Verify every **success criterion** from `{{roadmapPath}}` with evidence from summaries, tests, or observable behavior. Record unmet criteria as **verification failure**.
-5. Verify **definition of done**: all slices `[x]`, summaries exist, and integrations work. Record unmet items as **verification failure**.
+4. Verify every **success criterion** from `{{roadmapPath}}`. If passing validation is present, summarize the validation evidence instead of re-auditing it; otherwise verify with evidence from summaries, tests, or observable behavior. Record unmet criteria as **verification failure**.
+5. Verify **definition of done**: all slices `[x]`, summaries exist, and integrations work. If passing validation is present, trust its integration/verification verdict unless inconsistent with current artifacts. Record unmet items as **verification failure**.
 6. If the roadmap includes a **Horizontal Checklist**, verify each item and note unchecked items in the summary.
 7. Fill the **Decision Re-evaluation** table: compare each key `.gsd/DECISIONS.md` decision from this milestone with what shipped, and flag decisions to revisit.
 8. Validate **requirement status transitions**. For each changed requirement, confirm evidence supports the new status. Requirements may move between Active, Validated, Deferred, Blocked, or Out of Scope only with proof.

package/src/resources/extensions/gsd/prompts/plan-milestone.md

@@ -48,7 +48,7 @@ Narrate decomposition reasoning in complete sentences: grouping, risk order, ver
 Then:
 1. Use the **Roadmap** output template from the inlined context above
 2. {{skillActivation}}
-3. Create only as many demoable vertical slices as the work genuinely needs.
+3. Create only as many demoable vertical slices as the work genuinely needs. Use 1-10 slices, sized to the work; tiny/single-file/static work should usually be one slice.
 4. Order by risk, high-risk first.
 5. Call `gsd_plan_milestone` to persist milestone fields, slice rows, and **Horizontal Checklist** through the DB-backed path. Fill checklist concerns considered during planning: requirements, decisions, shutdown, revenue, auth, shared resources, reconnection. Omit for trivial milestones. Do **not** write `{{outputPath}}`, `ROADMAP.md`, or other planning artifacts manually; the tool owns rendering and persistence.
 6. If planning produced structural decisions (slice ordering, technology choices, scope exclusions), call `gsd_decision_save` for each; the tool assigns IDs and regenerates `.gsd/DECISIONS.md`.
@@ -78,6 +78,8 @@ Apply these when decomposing and ordering slices:
 - Ship features, not proofs; use clearly marked realistic stubs only when necessary.
 - **Dependency format is comma-separated, never range syntax.** Write `depends:[S01,S02,S03]`, not `depends:[S01-S03]`.
 - Roadmap ambition must match the milestone; right-size decomposition.
+- Missing ecosystem markers are not a reason to over-plan. If Project Classification says `untyped-existing`, treat the listed content files as the project surface and use generic file-level workflow guidance.
+- For `untyped-existing` projects with 1-2 content files, prefer exactly one slice unless the request clearly spans multiple independent user-visible capabilities. For 3-5 content files, prefer 1-2 slices.
 
 ## Progressive Planning (ADR-011)
 

package/src/resources/extensions/gsd/safety/evidence-collector.ts

@@ -50,6 +50,15 @@ export interface FileEditEvidence {
 
 export type EvidenceEntry = BashEvidence | FileWriteEvidence | FileEditEvidence;
 
+const EXECUTION_TOOL_NAMES = new Set([
+  "bash",
+  "Bash",
+  "gsd_exec",
+  "gsd_exec_search",
+  "mcp__gsd-workflow__gsd_exec",
+  "mcp__gsd-workflow__gsd_exec_search",
+]);
+
 // ─── Module State ───────────────────────────────────────────────────────────
 
 let unitEvidence: EvidenceEntry[] = [];
@@ -188,11 +197,11 @@ export function clearEvidenceFromDisk(
  * Exit codes and output are filled in by recordToolResult after execution.
  */
 export function recordToolCall(toolCallId: string, toolName: string, input: Record<string, unknown>): void {
-  if (toolName === "bash" || toolName === "Bash") {
+  if (EXECUTION_TOOL_NAMES.has(toolName)) {
     unitEvidence.push({
       kind: "bash",
       toolCallId,
-      command: String(input.command ?? ""),
+      command: String(input.command ?? input.cmd ?? input.query ?? ""),
       exitCode: -1,
       outputSnippet: "",
       timestamp: Date.now(),

package/src/resources/extensions/gsd/tests/auto-abort-pause-regression.test.ts

@@ -1,7 +1,7 @@
 import test from "node:test";
 import assert from "node:assert/strict";
 
-import { _buildAbortedPauseContext } from "../bootstrap/agent-end-recovery.js";
+import { _buildAbortedPauseContext, isUserInitiatedAbortMessage } from "../bootstrap/agent-end-recovery.js";
 import { _buildCancelledUnitStopReason } from "../auto/phases.js";
 
 test("aborted agent_end maps errorMessage into structured aborted pause context", () => {
@@ -30,3 +30,9 @@ test("cancelled non-session failures are labeled as unit aborts (not session-cre
   assert.equal(cancelled.stopReason, "Unit aborted: tool invocation cancelled");
   assert.equal(cancelled.loopReason, "unit-aborted");
 });
+
+test("provider user-abort errors are recognized as cancellations, not provider outages", () => {
+  assert.equal(isUserInitiatedAbortMessage("Claude Code process aborted by user"), true);
+  assert.equal(isUserInitiatedAbortMessage("Request aborted by user"), true);
+  assert.equal(isUserInitiatedAbortMessage("HTTP 503 Service Unavailable"), false);
+});

package/src/resources/extensions/gsd/tests/auto-dashboard.test.ts

@@ -23,6 +23,8 @@ import {
   _refreshLastCommitForTests,
   _getLastCommitForTests,
   _getLastCommitFetchedAtForTests,
+  formatRuntimeHealthSignal,
+  shouldRenderRoadmapProgress,
 } from "../auto-dashboard.ts";
 import {
   openDatabase,
@@ -193,6 +195,37 @@ test("formatWidgetTokens formats millions with M", () => {
   assert.equal(formatWidgetTokens(25_000_000), "25M");
 });
 
+test("formatRuntimeHealthSignal surfaces idle recovery instead of generic progress", () => {
+  const signal = formatRuntimeHealthSignal({
+    version: 1,
+    unitType: "research-milestone",
+    unitId: "M001",
+    startedAt: 1_000,
+    updatedAt: 600_000,
+    phase: "recovered",
+    wrapupWarningSent: false,
+    continueHereFired: false,
+    timeoutAt: null,
+    lastProgressAt: 1_000,
+    progressCount: 1,
+    lastProgressKind: "idle-recovery-retry",
+    recoveryAttempts: 1,
+    lastRecoveryReason: "idle",
+  }, 600_000);
+
+  assert.deepEqual(signal, {
+    level: "yellow",
+    summary: "Recovering",
+    detail: "retry 1 after idle stall",
+  });
+});
+
+test("shouldRenderRoadmapProgress hides pre-roadmap zero-slice progress", () => {
+  assert.equal(shouldRenderRoadmapProgress(null), false);
+  assert.equal(shouldRenderRoadmapProgress({ done: 0, total: 0, activeSliceTasks: null } as any), false);
+  assert.equal(shouldRenderRoadmapProgress({ done: 0, total: 1, activeSliceTasks: null } as any), true);
+});
+
 // ─── estimateTimeRemaining ──────────────────────────────────────────────
 
 test("estimateTimeRemaining returns null when no ledger data", () => {

package/src/resources/extensions/gsd/tests/auto-loop.test.ts

@@ -879,7 +879,11 @@ test("autoLoop passes structured session-lock failure details to the handler", a
   );
 });
 
-test("autoLoop keeps queued sidecar work when the session lock is lost", async () => {
+// Regression for #5308: the iteration prelude must dequeue sidecar items
+// (popping the queue and emitting the `sidecar-dequeue` journal event) BEFORE
+// validateSessionLock + break-on-invalid. Inverting that order silently drops
+// queued sidecar work on lock-loss. Covers first-iteration and mid-session.
+test("autoLoop dequeues sidecar item before session-lock break (first iteration, #5308)", async () => {
   _resetPendingResolve();
 
   const ctx = makeMockCtx();
@@ -911,12 +915,87 @@ test("autoLoop keeps queued sidecar work when the session lock is lost", async (
 
   await autoLoop(ctx, pi, s, deps);
 
-  assert.equal(s.sidecarQueue.length, 1, "lost session lock must not consume queued sidecar work");
-  assert.equal(s.sidecarQueue[0]?.unitId, "M001/S01/T01/review");
-  assert.ok(!journalEvents.includes("sidecar-dequeue"), "sidecar dequeue must happen only after lock validation");
+  assert.equal(
+    s.sidecarQueue.length,
+    0,
+    "sidecar item must be popped on lock-loss iteration (pre-#5308 ordering)",
+  );
+  assert.ok(
+    journalEvents.includes("sidecar-dequeue"),
+    "sidecar-dequeue journal event must be emitted before session-lock break",
+  );
+  assert.ok(
+    deps.callLog.includes("handleLostSessionLock"),
+    "session lock handler must still fire after sidecar dequeue",
+  );
   assert.ok(!deps.callLog.includes("deriveState"), "lock loss should stop before deriving state");
 });
 
+test("autoLoop dequeues sidecar item before session-lock break (mid-session, #5308)", async () => {
+  _resetPendingResolve();
+
+  const ctx = makeMockCtx();
+  ctx.ui.setStatus = () => {};
+  const pi = makeMockPi();
+  const s = makeLoopSession();
+
+  const journalEvents: string[] = [];
+  let lockCheckCount = 0;
+  const deps = makeMockDeps({
+    // First iteration: lock valid; second iteration: lock invalidates.
+    validateSessionLock: () => {
+      lockCheckCount += 1;
+      if (lockCheckCount === 1) {
+        return { valid: true } as SessionLockStatus;
+      }
+      return {
+        valid: false,
+        failureReason: "compromised",
+        expectedPid: process.pid,
+      } as SessionLockStatus;
+    },
+    handleLostSessionLock: () => {
+      deps.callLog.push("handleLostSessionLock");
+    },
+    emitJournalEvent: (entry) => {
+      journalEvents.push(entry.eventType);
+    },
+    // Enqueue a sidecar item at the end of iteration 1, so iteration 2 begins
+    // with a non-empty queue and an invalid lock.
+    postUnitPostVerification: async () => {
+      deps.callLog.push("postUnitPostVerification");
+      s.sidecarQueue.push({
+        kind: "hook" as const,
+        unitType: "hook/review",
+        unitId: "M001/S01/T01/review",
+        prompt: "review the code",
+      });
+      return "continue" as const;
+    },
+  });
+
+  const loopPromise = autoLoop(ctx, pi, s, deps);
+  // Allow the loop to reach runUnit's await on iteration 1.
+  await new Promise((r) => setTimeout(r, 50));
+  resolveAgentEnd(makeEvent());
+  await loopPromise;
+
+  assert.ok(lockCheckCount >= 2, "lock validator must run on iteration 2");
+  assert.equal(
+    s.sidecarQueue.length,
+    0,
+    "queued sidecar item must be popped on the lock-loss iteration",
+  );
+  assert.ok(
+    journalEvents.includes("sidecar-dequeue"),
+    "sidecar-dequeue journal event must be emitted before session-lock break",
+  );
+  assert.ok(
+    deps.callLog.includes("handleLostSessionLock"),
+    "lock-loss handler must still fire on iteration 2",
+  );
+});
+
 test("autoLoop exits on terminal blocked state", async (t) => {
   _resetPendingResolve();
 
@@ -2477,7 +2556,7 @@ test("autoLoop warns but proceeds for greenfield project (no project files) (#18
     "should not stop with health check failure for greenfield project",
   );
   const greenfieldWarning = notifications.find(
-    (n) => n.includes("no recognized project files") && n.includes("greenfield"),
+    (n) => n.includes("no project content yet") && n.includes("greenfield"),
   );
   assert.ok(
     greenfieldWarning,

package/src/resources/extensions/gsd/tests/auto-recovery.test.ts

@@ -7,11 +7,12 @@ import { randomUUID } from "node:crypto";
 
 import { verifyExpectedArtifact, hasImplementationArtifacts, resolveExpectedArtifactPath, diagnoseExpectedArtifact, buildLoopRemediationSteps, writeBlockerPlaceholder } from "../auto-recovery.ts";
 import { resolveMilestoneFile } from "../paths.ts";
-import { openDatabase, closeDatabase, insertMilestone, insertSlice, insertGateRow, insertTask } from "../gsd-db.ts";
+import { openDatabase, closeDatabase, insertMilestone, insertSlice, insertGateRow, insertTask, getMilestoneCommitAttributionShas } from "../gsd-db.ts";
 import { clearParseCache } from "../files.ts";
 import { parseRoadmap } from "../parsers-legacy.ts";
 import { invalidateAllCaches } from "../cache.ts";
 import { deriveState, invalidateStateCache } from "../state.ts";
+import { writeIntegrationBranch } from "../git-service.ts";
 
 const tmpDirs: string[] = [];
 
@@ -735,6 +736,174 @@ test("hasImplementationArtifacts rejects milestone-scoped main history with only
   }
 });
 
+test("hasImplementationArtifacts finds integration implementation-only commits when milestone branch diff is .gsd-only", () => {
+  const base = makeGitBase();
+  try {
+    mkdirSync(join(base, "src"), { recursive: true });
+    writeFileSync(join(base, "src", "feature.ts"), "export function feature() {}\n");
+    execFileSync("git", ["add", "src/feature.ts"], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: add milestone feature\n\nGSD-Task: S01/T01"], { cwd: base, stdio: "ignore" });
+
+    mkdirSync(join(base, ".gsd"), { recursive: true });
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      title: "Slice One",
+      status: "complete",
+      risk: "low",
+      depends: [],
+    });
+    insertTask({
+      id: "T01",
+      sliceId: "S01",
+      milestoneId: "M001",
+      title: "Task One",
+      status: "complete",
+    });
+
+    execFileSync("git", ["checkout", "-b", "milestone/M001"], { cwd: base, stdio: "ignore" });
+    writeIntegrationBranch(base, "M001", "main");
+    writeFileSync(join(base, ".gsd", "milestones", "M001", "M001-SUMMARY.md"), "# Milestone Summary\nDone.");
+    execFileSync("git", ["add", "."], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "chore: auto-commit after complete-milestone\n\nGSD-Unit: M001"], { cwd: base, stdio: "ignore" });
+
+    const result = hasImplementationArtifacts(base, "M001");
+    assert.equal(
+      result,
+      "present",
+      ".gsd-only milestone closeout diffs should still honor implementation commits already on the integration branch",
+    );
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("hasImplementationArtifacts backfills untagged main implementation commits from completed task file hints", () => {
+  const base = makeGitBase();
+  try {
+    mkdirSync(join(base, ".gsd"), { recursive: true });
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      title: "Slice One",
+      status: "complete",
+      risk: "low",
+      depends: [],
+    });
+    insertTask({
+      id: "T01",
+      sliceId: "S01",
+      milestoneId: "M001",
+      title: "Task One",
+      status: "complete",
+      keyFiles: ["index.html", "style.css", "app.js"],
+      planning: { files: ["index.html", "style.css", "app.js"] },
+    });
+
+    writeFileSync(join(base, "index.html"), "<main></main>\n");
+    writeFileSync(join(base, "style.css"), "main { display: block; }\n");
+    writeFileSync(join(base, "app.js"), "document.body.dataset.ready = 'true';\n");
+    execFileSync("git", ["add", "index.html", "style.css", "app.js"], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: add to-do app with CRUD and localStorage persistence"], { cwd: base, stdio: "ignore" });
+    const commitSha = execFileSync("git", ["rev-parse", "HEAD"], { cwd: base, encoding: "utf-8" }).trim();
+
+    const result = hasImplementationArtifacts(base, "M001");
+    assert.equal(
+      result,
+      "present",
+      "completed task file hints should repair prior untagged implementation commits on main",
+    );
+    assert.deepEqual(getMilestoneCommitAttributionShas("M001"), [commitSha]);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("hasImplementationArtifacts does not backfill untagged commits before milestone creation", () => {
+  const base = makeGitBase();
+  try {
+    writeFileSync(join(base, "app.js"), "document.body.dataset.ready = 'old';\n");
+    execFileSync("git", ["add", "app.js"], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: old app work"], {
+      cwd: base,
+      stdio: "ignore",
+      env: {
+        ...process.env,
+        GIT_AUTHOR_DATE: "2020-01-01T00:00:00Z",
+        GIT_COMMITTER_DATE: "2020-01-01T00:00:00Z",
+      },
+    });
+
+    mkdirSync(join(base, ".gsd"), { recursive: true });
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      title: "Slice One",
+      status: "complete",
+      risk: "low",
+      depends: [],
+    });
+    insertTask({
+      id: "T01",
+      sliceId: "S01",
+      milestoneId: "M001",
+      title: "Task One",
+      status: "complete",
+      keyFiles: ["app.js"],
+      planning: { files: ["app.js"] },
+    });
+
+    const result = hasImplementationArtifacts(base, "M001");
+    assert.equal(result, "absent", "pre-milestone commits must not be attributed to the milestone");
+    assert.deepEqual(getMilestoneCommitAttributionShas("M001"), []);
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("hasImplementationArtifacts does not backfill unrelated untagged implementation commits", () => {
+  const base = makeGitBase();
+  try {
+    mkdirSync(join(base, ".gsd"), { recursive: true });
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+    insertSlice({
+      id: "S01",
+      milestoneId: "M001",
+      title: "Slice One",
+      status: "complete",
+      risk: "low",
+      depends: [],
+    });
+    insertTask({
+      id: "T01",
+      sliceId: "S01",
+      milestoneId: "M001",
+      title: "Task One",
+      status: "complete",
+      keyFiles: ["src/expected.ts"],
+      planning: { files: ["src/expected.ts"] },
+    });
+
+    mkdirSync(join(base, "src"), { recursive: true });
+    writeFileSync(join(base, "src", "unrelated.ts"), "export const unrelated = true;\n");
+    execFileSync("git", ["add", "src/unrelated.ts"], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: unrelated work"], { cwd: base, stdio: "ignore" });
+
+    const result = hasImplementationArtifacts(base, "M001");
+    assert.equal(result, "absent", "backfill must require overlap with completed task file hints");
+    assert.deepEqual(getMilestoneCommitAttributionShas("M001"), []);
+  } finally {
+    cleanup(base);
+  }
+});
+
 test("hasImplementationArtifacts treats empty non-integration branch diff as absent (#4699)", () => {
   const base = makeGitBase();
   try {