gsd-pi 2.50.0 → 2.51.0-dev.7d435fe

package/src/resources/extensions/gsd/prompts/forensics.md

@@ -142,11 +142,10 @@ Then **offer GitHub issue creation**: "Would you like me to create a GitHub issu
 If yes, create using the `bash` tool:
 
 ```bash
-# Step 1: Create issue (use labels for metadata, NOT for classification — type is set via GraphQL)
-ISSUE_URL=$(gh issue create --repo gsd-build/gsd-2 \
-  --title "..." \
-  --label "auto-generated" \
-  --body "$(cat <<'EOF'
+# Step 1: Write issue body to a temp file to avoid escaping/truncation issues.
+# Using --body-file bypasses shell quoting entirely — backticks, quotes, and
+# content containing "EOF" all render correctly. (#2465)
+cat > /tmp/gsd-forensic-issue.md << 'GSD_ISSUE_BODY'
 ## Problem
 [1-2 sentence summary]
 
@@ -169,8 +168,13 @@ ISSUE_URL=$(gh issue create --repo gsd-build/gsd-2 \
 
 ---
 *Auto-generated by `/gsd forensics`*
-EOF
-)")
+GSD_ISSUE_BODY
+
+ISSUE_URL=$(gh issue create --repo gsd-build/gsd-2 \
+  --title "..." \
+  --label "auto-generated" \
+  --body-file /tmp/gsd-forensic-issue.md)
+rm -f /tmp/gsd-forensic-issue.md
 
 # Step 2: Set issue type via GraphQL (gh issue create has no --type flag)
 ISSUE_NUM=$(echo "$ISSUE_URL" | grep -oE '[0-9]+$')

package/src/resources/extensions/gsd/prompts/system.md

@@ -92,7 +92,7 @@ Titles live inside file content (headings, frontmatter), not in file or director
 
 ### Isolation Model
 
-Auto-mode supports three isolation modes (configured in `.gsd/preferences.md` under `taskIsolation.mode`):
+Auto-mode supports three isolation modes (configured in `.gsd/PREFERENCES.md` under `taskIsolation.mode`):
 
 - **worktree** (default): Work happens in `.gsd/worktrees/<MID>/`, a full git worktree on the `milestone/<MID>` branch. Each worktree has its own working copy and `.gsd/` directory. Squash-merged back to the integration branch on milestone completion.
 - **branch**: Work happens in the project root on a `milestone/<MID>` branch. No worktree directory — files are checked out in-place.

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

@@ -24,7 +24,12 @@ All relevant context has been preloaded below — the roadmap, all slice summari
 2. For each **slice** in the roadmap, verify its demo/deliverable claim against its summary. Flag any slice whose summary does not substantiate its claimed output.
 3. Check **cross-slice integration points** — do boundary map entries (produces/consumes) align with what was actually built?
 4. Check **requirement coverage** — are all active requirements addressed by at least one slice?
-5. Determine a verdict:
+5. If **Verification Classes** are provided in the inlined context above, check each non-empty class:
+   - For each verification class (Contract, Integration, Operational, UAT), determine whether slice summaries, UAT results, or observable behavior provide evidence that this verification tier was addressed.
+   - Document the compliance status of each class in your verdict rationale.
+   - If `Operational` verification is non-empty and no evidence of operational verification exists, flag this explicitly — it means planned operational checks (migrations, deployments, runtime verification) were not proven.
+   - A milestone with unaddressed verification classes may still pass if the gaps are minor, but the gaps MUST be documented in the Deferred Work Inventory.
+6. Determine a verdict:
    - `pass` — all criteria met, all slices delivered, no gaps
    - `needs-attention` — minor gaps that do not block completion (document them)
    - `needs-remediation` — material gaps found; remediation slices must be added to the roadmap

package/src/resources/extensions/gsd/provider-error-pause.ts

@@ -2,54 +2,6 @@ export type ProviderErrorPauseUI = {
   notify(message: string, level?: "info" | "warning" | "error" | "success"): void;
 };
 
-/**
- * Classify a provider error as transient (auto-resume) or permanent (manual resume).
- *
- * Transient: rate limits, server errors (500/502/503), overloaded, internal errors.
- * These are expected to self-resolve and should auto-resume after a delay.
- *
- * Permanent: auth errors, invalid API key, billing issues.
- * These require user intervention and should pause indefinitely.
- */
-export function classifyProviderError(errorMsg: string): {
-  isTransient: boolean;
-  isRateLimit: boolean;
-  suggestedDelayMs: number;
-} {
-  const isRateLimit = /rate.?limit|too many requests|429/i.test(errorMsg);
-  const isServerError = /internal server error|500|502|503|overloaded|server_error|api_error|service.?unavailable/i.test(errorMsg);
-
-  // Connection/process errors — transient, auto-resume after brief backoff (#2309).
-  // These indicate the process was killed, the connection was reset, or a network
-  // blip occurred. They are NOT permanent failures.
-  const isConnectionError = /terminated|connection.?reset|connection.?refused|other side closed|fetch failed|network.?(?:is\s+)?unavailable|ECONNREFUSED|ECONNRESET|EPIPE/i.test(errorMsg);
-
-  // Permanent errors — never auto-resume
-  const isPermanent = /auth|unauthorized|forbidden|invalid.*key|invalid.*api|billing|quota exceeded|account/i.test(errorMsg);
-
-  if (isPermanent && !isRateLimit) {
-    return { isTransient: false, isRateLimit: false, suggestedDelayMs: 0 };
-  }
-
-  if (isRateLimit) {
-    // Try to extract retry-after from the message
-    const resetMatch = errorMsg.match(/reset in (\d+)s/i);
-    const delayMs = resetMatch ? Number(resetMatch[1]) * 1000 : 60_000; // default 60s for rate limits
-    return { isTransient: true, isRateLimit: true, suggestedDelayMs: delayMs };
-  }
-
-  if (isServerError) {
-    return { isTransient: true, isRateLimit: false, suggestedDelayMs: 30_000 }; // 30s for server errors
-  }
-
-  if (isConnectionError) {
-    return { isTransient: true, isRateLimit: false, suggestedDelayMs: 15_000 }; // 15s for connection errors
-  }
-
-  // Unknown error — treat as permanent (user reviews)
-  return { isTransient: false, isRateLimit: false, suggestedDelayMs: 0 };
-}
-
 /**
  * Pause auto-mode due to a provider error.
  *

package/src/resources/extensions/gsd/roadmap-mutations.ts

@@ -0,0 +1,134 @@
+/**
+ * Roadmap Mutations — shared utilities for modifying roadmap checkbox state.
+ *
+ * Extracts the duplicated "flip slice checkbox" pattern that existed in
+ * doctor.ts, mechanical-completion.ts, and auto-recovery.ts.
+ */
+
+import { readFileSync } from "node:fs";
+import { atomicWriteSync } from "./atomic-write.js";
+import { resolveMilestoneFile } from "./paths.js";
+import { clearParseCache } from "./files.js";
+
+/**
+ * Mark a slice as done ([x]) in the milestone roadmap.
+ * Idempotent — no-op if already checked or if the slice isn't found.
+ *
+ * @returns true if the roadmap was modified, false if no change was needed
+ */
+export function markSliceDoneInRoadmap(basePath: string, mid: string, sid: string): boolean {
+  const roadmapFile = resolveMilestoneFile(basePath, mid, "ROADMAP");
+  if (!roadmapFile) return false;
+
+  let content: string;
+  try {
+    content = readFileSync(roadmapFile, "utf-8");
+  } catch {
+    return false;
+  }
+
+  // Try checkbox format first: "- [ ] **S01: Title**"
+  let updated = content.replace(
+    new RegExp(`^(\\s*-\\s+)\\[ \\]\\s+\\*\\*${sid}:`, "m"),
+    `$1[x] **${sid}:`,
+  );
+
+  // If checkbox format didn't match, try prose format: "## S01: Title" -> "## S01: \u2713 Title"
+  if (updated === content) {
+    updated = content.replace(
+      new RegExp(`^(#{1,4}\\s+(?:\\*{0,2})(?:Slice\\s+)?${sid}\\*{0,2}[:\\s.\\u2014\\u2013-]+\\s*)(.+)`, "m"),
+      (match, prefix, title) => {
+        // Already marked done — no-op
+        if (/^\u2713/.test(title) || /\(Complete\)\s*$/i.test(title)) return match;
+        return `${prefix}\u2713 ${title}`;
+      },
+    );
+  }
+
+  if (updated === content) return false;
+
+  atomicWriteSync(roadmapFile, updated);
+  clearParseCache();
+  return true;
+}
+
+/**
+ * Mark a slice as not done ([ ]) in the milestone roadmap.
+ * Idempotent — no-op if already unchecked or if the slice isn't found.
+ *
+ * @returns true if the roadmap was modified, false if no change was needed
+ */
+export function markSliceUndoneInRoadmap(basePath: string, mid: string, sid: string): boolean {
+  const roadmapFile = resolveMilestoneFile(basePath, mid, "ROADMAP");
+  if (!roadmapFile) return false;
+
+  let content: string;
+  try {
+    content = readFileSync(roadmapFile, "utf-8");
+  } catch {
+    return false;
+  }
+
+  const updated = content.replace(
+    new RegExp(`^(\\s*-\\s+)\\[x\\]\\s+\\*\\*${sid}:`, "m"),
+    `$1[ ] **${sid}:`,
+  );
+
+  if (updated === content) return false;
+
+  atomicWriteSync(roadmapFile, updated);
+  clearParseCache();
+  return true;
+}
+
+/**
+ * Mark a task as done ([x]) in the slice plan.
+ * Idempotent — no-op if already checked or if the task isn't found.
+ *
+ * @returns true if the plan was modified, false if no change was needed
+ */
+export function markTaskDoneInPlan(basePath: string, planPath: string, tid: string): boolean {
+  let content: string;
+  try {
+    content = readFileSync(planPath, "utf-8");
+  } catch {
+    return false;
+  }
+
+  const updated = content.replace(
+    new RegExp(`^(\\s*-\\s+)\\[ \\]\\s+\\*\\*${tid}:`, "m"),
+    `$1[x] **${tid}:`,
+  );
+
+  if (updated === content) return false;
+
+  atomicWriteSync(planPath, updated);
+  clearParseCache();
+  return true;
+}
+
+/**
+ * Mark a task as not done ([ ]) in the slice plan.
+ * Idempotent — no-op if already unchecked or if the task isn't found.
+ *
+ * @returns true if the plan was modified, false if no change was needed
+ */
+export function markTaskUndoneInPlan(basePath: string, planPath: string, tid: string): boolean {
+  let content: string;
+  try {
+    content = readFileSync(planPath, "utf-8");
+  } catch {
+    return false;
+  }
+
+  const updated = content.replace(
+    new RegExp(`^(\\s*-\\s+)\\[x\\]\\s+\\*\\*${tid}:`, "mi"),
+    `$1[ ] **${tid}:`,
+  );
+
+  if (updated === content) return false;
+
+  atomicWriteSync(planPath, updated);
+  clearParseCache();
+  return true;
+}

package/src/resources/extensions/gsd/rtk-status.ts

@@ -0,0 +1,53 @@
+import type { ExtensionContext } from "@gsd/pi-coding-agent";
+import {
+  ensureRtkSessionBaseline,
+  formatRtkSavingsLabel,
+  getRtkSessionSavings,
+} from "../shared/rtk-session-stats.js";
+import { loadEffectiveGSDPreferences } from "./preferences.js";
+
+const STATUS_KEY = "gsd-rtk";
+const REFRESH_INTERVAL_MS = 30_000;
+
+let refreshTimer: ReturnType<typeof setInterval> | null = null;
+
+function clearTimer(): void {
+  if (refreshTimer) {
+    clearInterval(refreshTimer);
+    refreshTimer = null;
+  }
+}
+
+function isRtkEnabledInPrefs(): boolean {
+  return loadEffectiveGSDPreferences()?.preferences.experimental?.rtk === true;
+}
+
+function updateStatus(ctx: ExtensionContext): void {
+  if (!ctx.hasUI) return;
+  if (!isRtkEnabledInPrefs()) return;
+
+  const basePath = ctx.cwd;
+  const sessionId = ctx.sessionManager.getSessionId();
+  ensureRtkSessionBaseline(basePath, sessionId);
+  const savings = getRtkSessionSavings(basePath, sessionId);
+  ctx.ui.setStatus(STATUS_KEY, formatRtkSavingsLabel(savings) ?? undefined);
+}
+
+export function startRtkStatusUpdates(ctx: ExtensionContext): void {
+  clearTimer();
+  if (!isRtkEnabledInPrefs()) {
+    // Ensure any previously set status is cleared (e.g. preference was toggled off)
+    ctx.ui.setStatus(STATUS_KEY, undefined);
+    return;
+  }
+  updateStatus(ctx);
+  if (!ctx.hasUI) return;
+  refreshTimer = setInterval(() => {
+    updateStatus(ctx);
+  }, REFRESH_INTERVAL_MS);
+}
+
+export function stopRtkStatusUpdates(ctx?: ExtensionContext): void {
+  clearTimer();
+  ctx?.ui.setStatus(STATUS_KEY, undefined);
+}

package/src/resources/extensions/gsd/rule-registry.ts

@@ -20,20 +20,19 @@ import type {
 import { resolvePostUnitHooks, resolvePreDispatchHooks } from "./preferences.js";
 import { existsSync, readFileSync, writeFileSync, mkdirSync } from "node:fs";
 import { join } from "node:path";
+import { parseUnitId } from "./unit-id.js";
 
 // ─── Artifact Path Resolution ──────────────────────────────────────────────
 
 export function resolveHookArtifactPath(basePath: string, unitId: string, artifactName: string): string {
-  const parts = unitId.split("/");
-  if (parts.length === 3) {
-    const [mid, sid, tid] = parts;
-    return join(basePath, ".gsd", "milestones", mid, "slices", sid, "tasks", `${tid}-${artifactName}`);
+  const { milestone, slice, task } = parseUnitId(unitId);
+  if (task !== undefined && slice !== undefined) {
+    return join(basePath, ".gsd", "milestones", milestone, "slices", slice, "tasks", `${task}-${artifactName}`);
   }
-  if (parts.length === 2) {
-    const [mid, sid] = parts;
-    return join(basePath, ".gsd", "milestones", mid, "slices", sid, artifactName);
+  if (slice !== undefined) {
+    return join(basePath, ".gsd", "milestones", milestone, "slices", slice, artifactName);
   }
-  return join(basePath, ".gsd", "milestones", parts[0], artifactName);
+  return join(basePath, ".gsd", "milestones", milestone, artifactName);
 }
 
 // ─── Dispatch Rule Conversion ──────────────────────────────────────────────
@@ -212,7 +211,7 @@ export class RuleRegistry {
       };
 
       // Build prompt with variable substitution
-      const [mid, sid, tid] = triggerUnitId.split("/");
+      const { milestone: mid, slice: sid, task: tid } = parseUnitId(triggerUnitId);
       let prompt = config.prompt
         .replace(/\{milestoneId\}/g, mid ?? "")
         .replace(/\{sliceId\}/g, sid ?? "")
@@ -291,7 +290,7 @@ export class RuleRegistry {
       return { action: "proceed", prompt, firedHooks: [] };
     }
 
-    const [mid, sid, tid] = unitId.split("/");
+    const { milestone: mid, slice: sid, task: tid } = parseUnitId(unitId);
     const substitute = (text: string): string =>
       text
         .replace(/\{milestoneId\}/g, mid ?? "")
@@ -506,7 +505,7 @@ export class RuleRegistry {
     this.cycleCounts.set(cycleKey, currentCycle);
     this.activeHook.cycle = currentCycle;
 
-    const [mid, sid, tid] = unitId.split("/");
+    const { milestone: mid, slice: sid, task: tid } = parseUnitId(unitId);
     const prompt = hook.prompt
       .replace(/\{milestoneId\}/g, mid ?? "")
       .replace(/\{sliceId\}/g, sid ?? "")
@@ -525,7 +524,7 @@ export class RuleRegistry {
   formatHookStatus(): string {
     const entries = this.getHookStatus();
     if (entries.length === 0) {
-      return "No hooks configured. Add post_unit_hooks or pre_dispatch_hooks to .gsd/preferences.md";
+      return "No hooks configured. Add post_unit_hooks or pre_dispatch_hooks to .gsd/PREFERENCES.md";
     }
 
     const lines: string[] = ["Configured Hooks:", ""];

package/src/resources/extensions/gsd/service-tier.ts

@@ -27,15 +27,27 @@ const SERVICE_TIER_SCOPE_NOTE = "Only affects gpt-5.4 models, regardless of prov
 
 // ─── Gating ──────────────────────────────────────────────────────────────────
 
+/**
+ * Model ID prefixes (bare, without provider) that support OpenAI service tiers.
+ *
+ * This list is the fallback for callers that only have a model ID string.
+ * The authoritative source of truth is `model.capabilities.supportsServiceTier`
+ * (set via CAPABILITY_PATCHES in packages/pi-ai/src/models.ts). When callers
+ * have access to the full Model object, prefer reading capabilities directly.
+ *
+ * See: https://github.com/gsd-build/gsd-2/issues/2546
+ */
+const SERVICE_TIER_MODEL_PREFIXES = ["gpt-5.4"] as const;
+
 /**
  * Returns true when the given model ID supports OpenAI service tiers.
- * Currently only gpt-5.4 variants qualify.
+ * Reads from SERVICE_TIER_MODEL_PREFIXES — update that list, not this function.
  */
 export function supportsServiceTier(modelId: string): boolean {
   if (!modelId) return false;
   // Strip provider prefix if present (e.g. "openai/gpt-5.4" → "gpt-5.4")
   const bare = modelId.includes("/") ? modelId.split("/").pop()! : modelId;
-  return bare.startsWith("gpt-5.4");
+  return SERVICE_TIER_MODEL_PREFIXES.some((prefix) => bare.startsWith(prefix));
 }
 
 // ─── Status Formatting ───────────────────────────────────────────────────────

package/src/resources/extensions/gsd/session-forensics.ts

@@ -472,9 +472,13 @@ function formatTraceSummary(trace: ExecutionTrace): string {
   if (trace.errors.length > 0) {
     parts.push(`Errors: ${trace.errors.slice(-3).join("; ")}`);
   }
-  if (trace.lastReasoning) {
-    parts.push(`Last reasoning: "${trace.lastReasoning}"`);
-  }
+  // NOTE: lastReasoning is intentionally excluded from the retry diagnostic.
+  // This summary is injected into retry prompts via getDeepDiagnostic() →
+  // phases.ts. Including prior assistant free-text causes hallucination loops
+  // when the previous turn was truncated or malformed. Crash recovery has its
+  // own path (formatCrashRecoveryBriefing) that handles lastReasoning safely
+  // with explicit "Last Agent Reasoning Before Interruption" framing.
+  // See: https://github.com/gsd-build/gsd-2/issues/2195
   return parts.join("\n");
 }