gsd-pi 2.35.0 → 2.36.0-dev.d612764

package/src/resources/extensions/gsd/tests/test-utils.ts

@@ -0,0 +1,165 @@
+/**
+ * Shared test utilities for GSD extension tests.
+ *
+ * Provides cross-platform helpers for creating temporary git repos,
+ * safe cleanup, file creation, and shell-free git operations.
+ *
+ * Usage:
+ *   import { git, makeTempRepo, cleanup, createFile } from "./test-utils.ts";
+ */
+
+import { execFileSync } from "node:child_process";
+import {
+  existsSync,
+  mkdirSync,
+  mkdtempSync,
+  readFileSync,
+  rmSync,
+  statSync,
+  writeFileSync,
+} from "node:fs";
+import { dirname, join } from "node:path";
+import { tmpdir } from "node:os";
+
+/**
+ * Shell-free git helper — uses execFileSync to bypass shell entirely.
+ * No quoting issues, no Windows cmd.exe incompatibilities.
+ *
+ * @param cwd - Working directory for git command
+ * @param args - Git arguments (e.g., "add", "-A")
+ * @returns trimmed stdout
+ */
+export function git(cwd: string, ...args: string[]): string {
+  return execFileSync("git", args, {
+    cwd,
+    encoding: "utf-8",
+    stdio: "pipe",
+  }).trim();
+}
+
+/**
+ * Create a temporary git repository with an initial commit.
+ * Configures user.email, user.name, and core.autocrlf=false for
+ * consistent behavior across platforms.
+ *
+ * @param prefix - Optional prefix for the temp directory name
+ * @returns absolute path to the temp repo
+ */
+export function makeTempRepo(prefix: string = "gsd-test-"): string {
+  const dir = mkdtempSync(join(tmpdir(), prefix));
+  git(dir, "init");
+  git(dir, "config", "user.email", "test@test.com");
+  git(dir, "config", "user.name", "Test");
+  git(dir, "config", "core.autocrlf", "false");
+  writeFileSync(join(dir, "README.md"), "# init\n");
+  git(dir, "add", "-A");
+  git(dir, "commit", "-m", "init");
+  git(dir, "branch", "-M", "main");
+  return dir;
+}
+
+/**
+ * Create a temporary directory (not a git repo).
+ *
+ * @param prefix - Optional prefix for the temp directory name
+ * @returns absolute path to the temp directory
+ */
+export function makeTempDir(prefix: string = "gsd-test-"): string {
+  return mkdtempSync(join(tmpdir(), prefix));
+}
+
+/**
+ * Safely clean up a temporary directory.
+ * Non-fatal — Windows may hold file descriptors briefly.
+ */
+export function cleanup(dir: string): void {
+  try {
+    rmSync(dir, { recursive: true, force: true });
+  } catch {
+    // ignore — Windows may hold file descriptors briefly after test
+  }
+}
+
+/**
+ * Create a file with intermediate directories.
+ *
+ * @param base - Base directory
+ * @param relativePath - Relative path within base (e.g., "src/index.ts")
+ * @param content - File content (defaults to empty string)
+ * @returns absolute path to the created file
+ */
+export function createFile(base: string, relativePath: string, content: string = ""): string {
+  const fullPath = join(base, relativePath);
+  mkdirSync(dirname(fullPath), { recursive: true });
+  writeFileSync(fullPath, content, "utf-8");
+  return fullPath;
+}
+
+/**
+ * Safely read a file, returning null if it doesn't exist or is a directory.
+ * Prevents EISDIR errors.
+ */
+export function safeReadFile(filePath: string): string | null {
+  try {
+    if (!existsSync(filePath)) return null;
+    if (!statSync(filePath).isFile()) return null;
+    return readFileSync(filePath, "utf-8");
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Create a minimal GSD milestone structure in a temp directory.
+ *
+ * @param base - Base directory (should have .gsd/ or be a temp repo)
+ * @param mid - Milestone ID (e.g., "M001")
+ * @param options - What to create
+ */
+export function writeMilestoneFixture(
+  base: string,
+  mid: string,
+  options: {
+    roadmap?: string;
+    context?: string;
+    summary?: string;
+    validation?: string;
+    slices?: Array<{
+      id: string;
+      plan?: string;
+      summary?: string;
+      uat?: string;
+    }>;
+  } = {},
+): void {
+  const milestoneDir = join(base, ".gsd", "milestones", mid);
+  mkdirSync(milestoneDir, { recursive: true });
+
+  if (options.roadmap) {
+    writeFileSync(join(milestoneDir, `${mid}-ROADMAP.md`), options.roadmap);
+  }
+  if (options.context) {
+    writeFileSync(join(milestoneDir, `${mid}-CONTEXT.md`), options.context);
+  }
+  if (options.summary) {
+    writeFileSync(join(milestoneDir, `${mid}-SUMMARY.md`), options.summary);
+  }
+  if (options.validation) {
+    writeFileSync(join(milestoneDir, `${mid}-VALIDATION.md`), options.validation);
+  }
+  if (options.slices) {
+    for (const slice of options.slices) {
+      const sliceDir = join(milestoneDir, "slices", slice.id);
+      mkdirSync(sliceDir, { recursive: true });
+      if (slice.plan) {
+        writeFileSync(join(sliceDir, `${slice.id}-PLAN.md`), slice.plan);
+      }
+      if (slice.summary) {
+        writeFileSync(join(sliceDir, `${slice.id}-SUMMARY.md`), slice.summary);
+      }
+      if (slice.uat) {
+        writeFileSync(join(sliceDir, `${slice.id}-UAT.md`), slice.uat);
+      }
+    }
+  }
+}

package/src/resources/extensions/gsd/tests/validate-directory.test.ts

@@ -101,6 +101,21 @@ test("validateDirectory: subdirectory of home is NOT blocked", () => {
   }
 });
 
+// Regression test for #1317: GSD worktree inside $HOME must not be blocked even
+// when the resolved project root equals $HOME (e.g. home dir is a git repo).
+test("validateDirectory: GSD worktree path nested under home is NOT blocked (#1317)", () => {
+  const worktreePath = join(homedir(), ".gsd", "worktrees", "M001");
+  mkdirSync(worktreePath, { recursive: true });
+  try {
+    // The worktree CWD itself is a valid location — it must pass.
+    const result = validateDirectory(worktreePath);
+    assert.equal(result.safe, true, "GSD worktree path should be safe to run in");
+    assert.equal(result.severity, "ok");
+  } finally {
+    rmSync(join(homedir(), ".gsd", "worktrees", "M001"), { recursive: true, force: true });
+  }
+});
+
 // ─── Temp directory root ─────────────────────────────────────────────────────────
 
 test("validateDirectory: temp directory root is blocked", () => {

package/src/resources/extensions/gsd/tests/validate-milestone.test.ts

@@ -104,6 +104,11 @@ test("isValidationTerminal returns true for verdict: needs-remediation (#832)",
   assert.equal(isValidationTerminal(content), true);
 });
 
+test("isValidationTerminal returns true for verdict: passed (#1429)", () => {
+  const content = "---\nverdict: passed\nremediation_round: 0\n---\n\n# Validation";
+  assert.equal(isValidationTerminal(content), true);
+});
+
 test("isValidationTerminal returns false for missing frontmatter", () => {
   const content = "# Validation\nNo frontmatter here.";
   assert.equal(isValidationTerminal(content), false);
@@ -196,6 +201,7 @@ test("dispatch rule matches validating-milestone phase", async () => {
   try {
     // Set up minimal milestone structure for the prompt builder
     writeRoadmap(base, "M001", ALL_DONE_ROADMAP);
+    writeSliceSummary(base, "M001", "S01", "# S01 Summary\nDone."); // Guard requires slice summaries (#1368)
 
     const ctx: DispatchContext = {
       basePath: base,
@@ -231,6 +237,7 @@ test("dispatch rule skips when skip_milestone_validation preference is set", asy
   const base = makeTmpBase();
   try {
     writeRoadmap(base, "M001", ALL_DONE_ROADMAP);
+    writeSliceSummary(base, "M001", "S01", "# S01 Summary\nDone."); // Guard requires slice summaries (#1368)
 
     const ctx: DispatchContext = {
       basePath: base,

package/src/resources/extensions/gsd/tests/worktree-sync-milestones.test.ts

@@ -19,6 +19,7 @@ import { join } from 'node:path';
 import { tmpdir } from 'node:os';
 
 import { syncProjectRootToWorktree } from '../auto-worktree-sync.ts';
+import { syncGsdStateToWorktree } from '../auto-worktree.ts';
 import { createTestContext } from './test-helpers.ts';
 
 const { assertTrue, report } = createTestContext();
@@ -148,6 +149,37 @@ async function main(): Promise<void> {
     assertTrue(true, 'no crash on missing directories');
   }
 
+  // ─── 7. milestones/ directory created in worktree when missing ────────
+  console.log('\n=== 7. milestones/ directory created in worktree when missing ===');
+  {
+    const mainBase = createBase('main');
+    const wtBase = mkdtempSync(join(tmpdir(), 'gsd-wt-sync-wt-'));
+
+    try {
+      // Worktree has .gsd/ but NO milestones/ subdirectory
+      mkdirSync(join(wtBase, '.gsd'), { recursive: true });
+
+      // Main repo has M001
+      const m001Dir = join(mainBase, '.gsd', 'milestones', 'M001');
+      mkdirSync(m001Dir, { recursive: true });
+      writeFileSync(join(m001Dir, 'M001-CONTEXT.md'), '# M001 Context');
+      writeFileSync(join(m001Dir, 'M001-ROADMAP.md'), '# M001 Roadmap');
+
+      assertTrue(!existsSync(join(wtBase, '.gsd', 'milestones')), 'milestones/ missing before sync');
+
+      const result = syncGsdStateToWorktree(mainBase, wtBase);
+
+      assertTrue(existsSync(join(wtBase, '.gsd', 'milestones')), 'milestones/ created in worktree');
+      assertTrue(existsSync(join(wtBase, '.gsd', 'milestones', 'M001')), 'M001 synced to worktree');
+      assertTrue(existsSync(join(wtBase, '.gsd', 'milestones', 'M001', 'M001-CONTEXT.md')), 'M001 CONTEXT synced');
+      assertTrue(existsSync(join(wtBase, '.gsd', 'milestones', 'M001', 'M001-ROADMAP.md')), 'M001 ROADMAP synced');
+      assertTrue(result.synced.length > 0, 'sync reported files');
+    } finally {
+      cleanup(mainBase);
+      rmSync(wtBase, { recursive: true, force: true });
+    }
+  }
+
   report();
 }
 

package/src/resources/extensions/gsd/worktree-resolver.ts

@@ -13,6 +13,8 @@
  * `process.chdir()` internally — this class MUST NOT double-chdir.
  */
 
+import { existsSync, unlinkSync } from "node:fs";
+import { join } from "node:path";
 import type { AutoSession } from "./auto/session.js";
 import { debugLog } from "./debug-logger.js";
 
@@ -372,6 +374,15 @@ export class WorktreeResolver {
       });
       ctx.notify(`Milestone merge failed: ${msg}`, "warning");
 
+      // Clean up stale merge state left by failed squash-merge (#1389)
+      try {
+        const gitDir = join(originalBase || this.s.basePath, ".git");
+        for (const f of ["SQUASH_MSG", "MERGE_HEAD", "MERGE_MSG"]) {
+          const p = join(gitDir, f);
+          if (existsSync(p)) unlinkSync(p);
+        }
+      } catch { /* best-effort */ }
+
       // Error recovery: always restore to project root
       if (originalBase) {
         try {

package/src/resources/extensions/remote-questions/remote-command.ts

@@ -4,12 +4,12 @@
 
 import type { ExtensionAPI, ExtensionCommandContext } from "@gsd/pi-coding-agent";
 import { AuthStorage } from "@gsd/pi-coding-agent";
-import { CURSOR_MARKER, Editor, type EditorTheme, Key, matchesKey, truncateToWidth } from "@gsd/pi-tui";
+import { Editor, type EditorTheme, Key, matchesKey, truncateToWidth } from "@gsd/pi-tui";
 import { existsSync, readFileSync, writeFileSync, mkdirSync } from "node:fs";
 import { dirname, join } from "node:path";
 import { getGlobalGSDPreferencesPath, loadEffectiveGSDPreferences } from "../gsd/preferences.js";
 import { getRemoteConfigStatus, isValidChannelId, resolveRemoteConfig } from "./config.js";
-import { sanitizeError } from "../shared/sanitize.js";
+import { maskEditorLine, sanitizeError } from "../shared/mod.js";
 import { getLatestPromptSummary } from "./status.js";
 
 export async function handleRemote(
@@ -353,27 +353,6 @@ function removeRemoteQuestionsConfig(): void {
   writeFileSync(prefsPath, next, "utf-8");
 }
 
-function maskEditorLine(line: string): string {
-  let output = "";
-  let i = 0;
-  while (i < line.length) {
-    if (line.startsWith(CURSOR_MARKER, i)) {
-      output += CURSOR_MARKER;
-      i += CURSOR_MARKER.length;
-      continue;
-    }
-    const ansiMatch = /^\x1b\[[0-9;]*m/.exec(line.slice(i));
-    if (ansiMatch) {
-      output += ansiMatch[0];
-      i += ansiMatch[0].length;
-      continue;
-    }
-    output += line[i] === " " ? " " : "*";
-    i += 1;
-  }
-  return output;
-}
-
 async function promptMaskedInput(ctx: ExtensionCommandContext, label: string, hint: string): Promise<string | null> {
   if (!ctx.hasUI) return null;
   return ctx.ui.custom<string | null>((tui: any, theme: any, _kb: any, done: (r: string | null) => void) => {

package/src/resources/extensions/search-the-web/native-search.ts

@@ -16,6 +16,16 @@ export const CUSTOM_SEARCH_TOOL_NAMES = ["search-the-web", "search_and_read", "g
 /** Thinking block types that require signature validation by the API */
 const THINKING_TYPES = new Set(["thinking", "redacted_thinking"]);
 
+/**
+ * Maximum number of native web searches allowed per session (agent unit).
+ * The Anthropic API's `max_uses` is per-request — it resets on each API call.
+ * When `pause_turn` triggers a resubmit, the model gets a fresh budget.
+ * This session-level cap prevents unbounded search accumulation (#1309).
+ *
+ * 15 = 3 full turns of 5 searches each — generous for research, but bounded.
+ */
+export const MAX_NATIVE_SEARCHES_PER_SESSION = 15;
+
 /** When true, skip native web search injection and keep Brave/custom tools active on Anthropic. */
 export function preferBraveSearch(): boolean {
   // preferences.md takes priority over env var
@@ -74,6 +84,11 @@ export function registerNativeSearchHooks(pi: NativeSearchPI): { getIsAnthropic:
   let isAnthropicProvider = false;
   let modelSelectFired = false;
 
+  // Session-level native search counter (#1309).
+  // Tracks cumulative web_search_tool_result blocks across all turns in a session.
+  // Reset on session_start. Used to compute remaining budget for max_uses.
+  let sessionSearchCount = 0;
+
   // Track provider changes via model selection — also handles diagnostics
   // since model_select fires AFTER session_start and knows the provider.
   pi.on("model_select", async (event: any, ctx: any) => {
@@ -161,13 +176,41 @@ export function registerNativeSearchHooks(pi: NativeSearchPI): { getIsAnthropic:
     );
     payload.tools = tools;
 
+    // ── Session-level search budget (#1309) ──────────────────────────────
+    // Count web_search_tool_result blocks in the conversation history to
+    // determine how many native searches have already been used this session.
+    // The Anthropic API's max_uses resets per request, so without this guard,
+    // pause_turn → resubmit cycles allow unlimited total searches.
+    if (Array.isArray(messages)) {
+      let historySearchCount = 0;
+      for (const msg of messages) {
+        const content = msg.content;
+        if (!Array.isArray(content)) continue;
+        for (const block of content) {
+          if ((block as any)?.type === "web_search_tool_result") {
+            historySearchCount++;
+          }
+        }
+      }
+      // Sync counter from history (handles session restore / context replay)
+      sessionSearchCount = historySearchCount;
+    }
+
+    const remaining = Math.max(0, MAX_NATIVE_SEARCHES_PER_SESSION - sessionSearchCount);
+
+    if (remaining <= 0) {
+      // Budget exhausted — don't inject the search tool at all.
+      // The model will proceed without web search capability.
+      return payload;
+    }
+
     tools.push({
       type: "web_search_20250305",
       name: "web_search",
-      // Cap server-side searches per response to prevent the model from
-      // looping on web_search without synthesizing results (#817).
-      // 5 searches is generous — most queries need 1-2.
-      max_uses: 5,
+      // Cap per-request searches to the lesser of 5 (per-turn cap) or the
+      // remaining session budget (#1309). This prevents the model from
+      // consuming unlimited searches via pause_turn → resubmit cycles.
+      max_uses: Math.min(5, remaining),
     });
 
     return payload;
@@ -175,6 +218,9 @@ export function registerNativeSearchHooks(pi: NativeSearchPI): { getIsAnthropic:
 
   // Basic startup diagnostics — provider-specific info comes from model_select
   pi.on("session_start", async (_event: any, ctx: any) => {
+    // Reset session-level search budget (#1309)
+    sessionSearchCount = 0;
+
     const hasBrave = !!process.env.BRAVE_API_KEY;
     const hasJina = !!process.env.JINA_API_KEY;
     const hasAnswers = !!process.env.BRAVE_ANSWERS_KEY;

package/src/resources/extensions/shared/mod.ts

@@ -28,6 +28,6 @@ export { showInterviewRound } from "./interview-ui.js";
 export type { Question, QuestionOption, RoundResult } from "./interview-ui.js";
 export { showNextAction } from "./next-action-ui.js";
 export { showConfirm } from "./confirm-ui.js";
-export { sanitizeError } from "./sanitize.js";
+export { sanitizeError, maskEditorLine } from "./sanitize.js";
 export { formatDateShort, truncateWithEllipsis } from "./format-utils.js";
 export { splitFrontmatter, parseFrontmatterMap } from "./frontmatter.js";

package/src/resources/extensions/shared/sanitize.ts

@@ -1,7 +1,10 @@
 /**
  * Sanitize error messages by redacting token-like strings before surfacing.
+ * Also provides maskEditorLine for masking sensitive TUI editor input.
  */
 
+import { CURSOR_MARKER } from "@gsd/pi-tui";
+
 const TOKEN_PATTERNS = [
   /xoxb-[A-Za-z0-9\-]+/g,    // Slack bot tokens
   /xoxp-[A-Za-z0-9\-]+/g,    // Slack user tokens
@@ -17,3 +20,36 @@ export function sanitizeError(msg: string): string {
   }
   return sanitized;
 }
+
+/**
+ * Replace editor visible text with masked characters while preserving
+ * ANSI cursor/sequencer codes. Keeps border/metadata lines readable.
+ */
+export function maskEditorLine(line: string): string {
+  if (line.startsWith("─")) {
+    return line;
+  }
+
+  let output = "";
+  let i = 0;
+  while (i < line.length) {
+    if (line.startsWith(CURSOR_MARKER, i)) {
+      output += CURSOR_MARKER;
+      i += CURSOR_MARKER.length;
+      continue;
+    }
+
+    const ansiMatch = /^\x1b\[[0-9;]*m/.exec(line.slice(i));
+    if (ansiMatch) {
+      output += ansiMatch[0];
+      i += ansiMatch[0].length;
+      continue;
+    }
+
+    const ch = line[i] as string;
+    output += ch === " " ? " " : "*";
+    i += 1;
+  }
+
+  return output;
+}

package/src/resources/extensions/shared/terminal.ts

@@ -7,9 +7,14 @@
 
 const UNSUPPORTED_TERMS = ["apple_terminal", "warpterm"];
 
+export function isCmuxTerminal(env: NodeJS.ProcessEnv = process.env): boolean {
+  return Boolean(env.CMUX_WORKSPACE_ID && env.CMUX_SURFACE_ID);
+}
+
 export function supportsCtrlAltShortcuts(): boolean {
   const term = (process.env.TERM_PROGRAM || "").toLowerCase();
   const jetbrains = (process.env.TERMINAL_EMULATOR || "").toLowerCase().includes("jetbrains");
+  if (isCmuxTerminal()) return true;
   return !UNSUPPORTED_TERMS.some((t) => term.includes(t)) && !jetbrains;
 }