@opengsd/gsd-pi 1.3.0 → 1.4.0

package/dist/resources/extensions/gsd/safety/evidence-collector.js

@@ -27,10 +27,14 @@ const EXECUTION_TOOL_NAMES = new Set([
 const MCP_EXECUTION_TOOL_RE = /^mcp__.+__gsd_(?:uat_)?exec(?:_search)?$/;
 // ─── Module State ───────────────────────────────────────────────────────────
 let unitEvidence = [];
+let lastWritePath = null;
+let lastWriteSig = null;
 // ─── Public API ─────────────────────────────────────────────────────────────
 /** Reset all evidence for a new unit. Call at unit start. */
 export function resetEvidence() {
     unitEvidence = [];
+    lastWritePath = null;
+    lastWriteSig = null;
 }
 /** Get a read-only view of all evidence collected for the current unit. */
 export function getEvidence() {
@@ -95,10 +99,15 @@ function isEvidenceArray(data) {
 export function saveEvidenceToDisk(basePath, milestoneId, sliceId, taskId) {
     try {
         const path = evidencePath(basePath, milestoneId, sliceId, taskId);
+        const body = JSON.stringify(unitEvidence, null, 2) + "\n";
+        if (path === lastWritePath && body === lastWriteSig)
+            return;
         mkdirSync(dirname(path), { recursive: true });
         const tmp = `${path}.tmp.${randomBytes(4).toString("hex")}`;
-        writeFileSync(tmp, JSON.stringify(unitEvidence, null, 2) + "\n", "utf-8");
+        writeFileSync(tmp, body, "utf-8");
         renameSync(tmp, path);
+        lastWritePath = path;
+        lastWriteSig = body;
     }
     catch {
         // Non-fatal — don't let persistence failures break unit execution
@@ -112,6 +121,8 @@ export function saveEvidenceToDisk(basePath, milestoneId, sliceId, taskId) {
  */
 export function loadEvidenceFromDisk(basePath, milestoneId, sliceId, taskId) {
     try {
+        lastWritePath = null;
+        lastWriteSig = null;
         const path = evidencePath(basePath, milestoneId, sliceId, taskId);
         if (!existsSync(path))
             return;
@@ -136,6 +147,10 @@ export function clearEvidenceFromDisk(basePath, milestoneId, sliceId, taskId) {
         if (existsSync(path)) {
             unlinkSync(path);
         }
+        if (path === lastWritePath) {
+            lastWritePath = null;
+            lastWriteSig = null;
+        }
     }
     catch {
         // Non-fatal

package/dist/resources/extensions/gsd/safety/evidence-cross-ref.js

@@ -38,8 +38,26 @@ export function crossReferenceEvidence(claimedEvidence, actualEvidence) {
             });
             continue;
         }
+        // A shell-spawn/infra failure means the command never ran (e.g. on Windows
+        // `gsd_exec runtime=bash` resolves to a WSL with no /bin/bash). The LLM
+        // typically recovers by re-running via another runtime, and that genuine
+        // run may not even be in the match set. Such a failure is inconclusive, not
+        // a falsified pass — exclude it before judging the exit code.
+        const commandRuns = matches.filter((m) => !isInfraSpawnFailure(m));
+        if (commandRuns.length === 0) {
+            if (claimed.exitCode === 0) {
+                mismatches.push({
+                    severity: "warning",
+                    claimed,
+                    actual: latestMatch(matches),
+                    reason: `Matched execution failed to spawn (infrastructure error, not a command failure); ` +
+                        `treating as inconclusive`,
+                });
+            }
+            continue;
+        }
         // Exit code mismatch: LLM claims success but actual command failed
-        const match = latestMatch(matches);
+        const match = latestMatch(commandRuns);
         if (claimed.exitCode === 0 && match.exitCode !== 0) {
             mismatches.push({
                 severity: "error",
@@ -51,6 +69,30 @@ export function crossReferenceEvidence(claimedEvidence, actualEvidence) {
     }
     return mismatches;
 }
+/**
+ * Runtime-spawn / shell-infra failure signatures. When a bash-runtime call
+ * fails to *spawn* (rather than the command running and exiting non-zero), the
+ * recorded exitCode is an ordinary non-zero but the output carries one of these
+ * markers. Such a call is not evidence that a verification failed.
+ */
+const INFRA_SPAWN_FAILURE_SIGNATURES = [
+    /execvpe\([^)]*\)\s+failed/i, // WSL: execvpe(/bin/bash) failed: No such file or directory
+    /WSL \(.*\) ERROR/i, // WSL relay error banner
+    /command not found:\s*(?:bash|sh|zsh|dash|fish|ash|ksh|wsl)\b/i, // missing shell interpreter only
+];
+/**
+ * True when a non-zero bash call looks like a shell-spawn/infra failure (the
+ * command never started) rather than a real command failure. A successful run
+ * (exitCode 0) is never an infra failure.
+ */
+function isInfraSpawnFailure(call) {
+    if (call.exitCode === 0)
+        return false;
+    const snippet = call.outputSnippet ?? "";
+    if (snippet.length === 0)
+        return false;
+    return INFRA_SPAWN_FAILURE_SIGNATURES.some((re) => re.test(snippet));
+}
 // ─── Internals ──────────────────────────────────────────────────────────────
 /**
  * Verification evidence rows are append-only across retries, but a task

package/dist/resources/extensions/gsd/safety/file-change-validator.js

@@ -44,8 +44,11 @@ export function validateFileChanges(basePath, expectedOutput, plannedFiles, file
     const actualFiles = getChangedFilesFromLastCommit(basePath);
     if (!actualFiles)
         return null;
-    // Filter out .gsd/ internal files — only validate project source files
-    const projectFiles = actualFiles.filter(f => !f.startsWith(".gsd/") && !f.startsWith(".gsd\\"));
+    // Filter out GSD-internal paths — only validate project source files.
+    const projectFiles = actualFiles.filter((f) => !f.startsWith(".gsd/") &&
+        !f.startsWith(".gsd\\") &&
+        !f.startsWith(".gsd-backups/") &&
+        !f.startsWith(".gsd-backups\\"));
     // Normalize expected paths (strip leading ./ or /)
     const normalizedExpected = new Set([...allExpected].map((f) => normalizePlannedFileReference(f).replace(/^\.\//, "").replace(/^\//, "")));
     // Build allowlist matchers once (dot: true so patterns like `**/.hidden` work).

package/dist/resources/extensions/gsd/service-tier.js

@@ -9,7 +9,7 @@
  */
 import { existsSync, readFileSync } from "node:fs";
 import { saveFile } from "./files.js";
-import { getGlobalGSDPreferencesPath, loadEffectiveGSDPreferences, loadGlobalGSDPreferences, } from "./preferences.js";
+import { clearGSDPreferencesCache, getGlobalGSDPreferencesPath, loadEffectiveGSDPreferences, loadGlobalGSDPreferences, } from "./preferences.js";
 import { ensurePreferencesFile, serializePreferencesToFrontmatter } from "./commands-prefs-wizard.js";
 const SERVICE_TIER_SCOPE_NOTE = "Only affects gpt-5.4 models, regardless of provider.";
 // ─── Gating ──────────────────────────────────────────────────────────────────
@@ -125,6 +125,7 @@ async function writeGlobalServiceTier(ctx, tier) {
             body = preserved;
     }
     await saveFile(path, `---\n${frontmatter}---${body}`);
+    clearGSDPreferencesCache();
     await ctx.waitForIdle();
     await ctx.reload();
 }

package/dist/resources/extensions/gsd/session-forensics.js

@@ -176,11 +176,21 @@ export function extractTrace(entries) {
     }
     // Flush any pending tool calls that never got results (crash mid-tool)
     for (const [, pending] of pendingTools) {
+        const missingResultError = `Tool call ${pending.name} started but no toolResult was recorded`;
         toolCalls.push({
             name: pending.name,
             input: redactInput(pending.name, pending.input),
-            isError: false,
+            result: "missing tool result (stream/tool-call abort before execution)",
+            isError: true,
         });
+        errors.push(missingResultError);
+        // Mark the matching commandsRun entry as failed so it is consistent with
+        // the isError: true on the tool call above (bash/bg_shell only).
+        if (pending.name === "bash" || pending.name === "bg_shell") {
+            const lastCmd = findLast(commandsRun, c => c.command === String(pending.input.command));
+            if (lastCmd)
+                lastCmd.failed = true;
+        }
     }
     return {
         toolCalls,

package/dist/resources/extensions/gsd/session-lock.js

@@ -604,6 +604,12 @@ function isLockFileOwnedByCurrentProcess(lp) {
     const existing = readExistingLockData(lp);
     return existing?.pid === process.pid;
 }
+const _retrySleepView = new Int32Array(new SharedArrayBuffer(4));
+function sleepSync(ms) {
+    if (!(ms > 0))
+        return;
+    Atomics.wait(_retrySleepView, 0, 0, ms);
+}
 export function readExistingLockDataWithRetry(lp, options) {
     const maxAttempts = options?.maxAttempts ?? 3;
     const delayMs = options?.delayMs ?? 200;
@@ -612,12 +618,7 @@ export function readExistingLockDataWithRetry(lp, options) {
         if (data !== null)
             return data;
         if (attempt < maxAttempts) {
-            // Synchronous busy-wait — onCompromised runs in a sync callback context
-            // and the delays are short (200ms default).
-            const start = Date.now();
-            while (Date.now() - start < delayMs) {
-                // busy-wait
-            }
+            sleepSync(delayMs);
         }
     }
     return null;

package/dist/resources/extensions/gsd/skills/gsd-headless/SKILL.md

@@ -240,3 +240,6 @@ Quick reference — see [references/commands.md](references/commands.md) for the
 | `skip` / `undo` | Unit control |
 | `doctor` | Health check + auto-fix |
 | `steer <desc>` | Hard-steer plan mid-execution |
+| `explore` / `spike` / `sketch` | Prompt-driven ideation and experiment workflows |
+| `map-codebase` / `docs-update` / `graphify` | Codebase documentation and knowledge workflows |
+| `code-review` / `review` / `audit-*` / `verify-work` | Review, audit, and UAT workflows |

package/dist/resources/extensions/gsd/skills/gsd-headless/references/commands.md

@@ -50,6 +50,57 @@ All commands can be run via `gsd headless [command]`.
 | `inspect` | Show SQLite DB diagnostics (schema, row counts) |
 | `forensics` | Post-mortem investigation of auto-mode failures |
 
+## Additional Prompt-Driven Workflows
+
+These commands dispatch native GSD prompts adapted to the milestone, slice, and `.gsd/` model. Use them as `gsd headless <command> [args...]`.
+
+| Command | Description |
+|---------|-------------|
+| `explore` | Socratic ideation before committing an idea |
+| `spike` | Focused throwaway experiment (`--quick`, `--text`, or frontier mode) |
+| `sketch` | UI/design exploration with throwaway HTML mockups |
+| `map-codebase` | Generate structured codebase docs under `.gsd/codebase/` |
+| `docs-update` | Generate, update, or verify docs against live code |
+| `graphify` | Build, query, inspect, or diff `.gsd/knowledge/` |
+| `stats` | Show project statistics, milestone state, git metrics, and timeline |
+| `progress` | Summarize recent work; can route `--next` or `--do "..."` |
+| `health` | Check `.gsd/` integrity (`--repair`, `--context`) |
+| `surface` | Manage surfaced skills and extensions |
+| `code-review` | Review changed source diff-first for bugs, security, and quality |
+| `review` | Peer-review recent work across reviewer perspectives |
+| `audit-milestone` | Verify a milestone met its definition of done |
+| `audit-uat` | Audit outstanding UAT/verification items |
+| `audit-fix` | Classify and remediate audit findings |
+| `ui-review` | Run a six-pillar frontend visual audit |
+| `secure-phase` | Verify threat mitigations |
+| `validate-phase` | Fill validation and test coverage gaps |
+| `verify-work` | Run conversational UAT |
+| `plan-review-convergence` | Iterate a plan through review cycles |
+| `discuss-phase` | Gather milestone or slice context through questions |
+| `plan-phase` | Create a detailed slice plan |
+| `execute-phase` | Execute slice tasks with wave support |
+| `spec-phase` | Clarify what a milestone delivers |
+| `mvp-phase` | Plan a vertical MVP milestone |
+| `ui-phase` | Produce a `UI-SPEC` |
+| `ai-integration-phase` | Produce an `AI-SPEC` |
+| `ultraplan-phase` | Run extended planning and review |
+| `autonomous` | Run remaining lifecycle work continuously |
+| `pause-work` | Create a pause handoff |
+| `resume-work` | Resume work with restored context |
+| `manager` | Manage multiple milestones from a command-center workflow |
+| `phase` | Manage milestone queue ordering |
+| `thread` | Manage persistent context threads |
+| `workstreams` | Route workstream actions through `parallel` |
+| `workspace` | Route workspace actions through `worktree` |
+| `milestone-summary` | Generate a project or milestone summary |
+| `review-backlog` | Review and promote backlog items |
+| `inbox` | Triage GitHub issues and PRs |
+| `import` | Ingest external plans with conflict detection |
+| `ingest-docs` | Bootstrap or merge `.gsd/` state from docs |
+| `profile-user` | Generate and persist a developer profile |
+| `settings` | Configure workflow toggles and model profile |
+| `ns-context`, `ns-ideate`, `ns-manage`, `ns-project`, `ns-review`, `ns-workflow` | Namespace-grouping names that redirect to `help` |
+
 ## Phases
 
 GSD workflows progress through these phases:

package/dist/resources/extensions/gsd/state/derive/cache.js

@@ -0,0 +1,28 @@
+// Project/App: gsd-pi
+// File Purpose: deriveState memoization cache and telemetry.
+const CACHE_TTL_MS = 100;
+let _stateCache = null;
+let _telemetry = { dbDeriveCount: 0 };
+export function getDeriveTelemetry() {
+    return { ..._telemetry };
+}
+export function resetDeriveTelemetry() {
+    _telemetry = { dbDeriveCount: 0 };
+}
+export function incrementDbDeriveCount() {
+    _telemetry.dbDeriveCount++;
+}
+export function invalidateStateCache() {
+    _stateCache = null;
+}
+export function readCachedDeriveState(cacheKey) {
+    if (_stateCache &&
+        _stateCache.basePath === cacheKey &&
+        Date.now() - _stateCache.timestamp < CACHE_TTL_MS) {
+        return _stateCache.result;
+    }
+    return null;
+}
+export function writeCachedDeriveState(cacheKey, result) {
+    _stateCache = { basePath: cacheKey, result, timestamp: Date.now() };
+}

package/dist/resources/extensions/gsd/state/derive/db-open.js

@@ -0,0 +1,39 @@
+// Project/App: gsd-pi
+// File Purpose: Workflow DB open helpers for state derivation.
+import { getAllMilestones, isDbAvailable, setMilestoneQueueOrder } from '../../gsd-db.js';
+import { openExistingWorkflowDatabase } from '../../db-workspace.js';
+import { loadQueueOrder, sortByQueueOrder } from '../../queue-order.js';
+export function syncQueueOrderProjectionToDb(basePath) {
+    const queueOrder = loadQueueOrder(basePath);
+    if (!queueOrder)
+        return;
+    const currentIds = getAllMilestones().map((m) => m.id);
+    const desiredIds = sortByQueueOrder(currentIds, queueOrder);
+    if (currentIds.length === desiredIds.length && currentIds.every((id, i) => id === desiredIds[i]))
+        return;
+    setMilestoneQueueOrder(desiredIds);
+}
+export function ensureExistingWorkflowDbOpen(basePath) {
+    const opened = isDbAvailable() || openExistingWorkflowDatabase(basePath).ok;
+    if (opened)
+        syncQueueOrderProjectionToDb(basePath);
+    return opened;
+}
+export function buildDbUnavailableState() {
+    return {
+        activeMilestone: null,
+        activeSlice: null,
+        activeTask: null,
+        phase: "pre-planning",
+        recentDecisions: [],
+        blockers: ["DB unavailable — runtime markdown state derivation is disabled"],
+        nextAction: "Open or create the canonical GSD database before deriving workflow state. If this project only has markdown state, run /gsd migrate explicitly.",
+        registry: [],
+        requirements: { active: 0, validated: 0, deferred: 0, outOfScope: 0, blocked: 0, total: 0 },
+        progress: { milestones: { done: 0, total: 0 } },
+    };
+}
+export function getRequestedMilestoneLock() {
+    const lock = process.env.GSD_MILESTONE_LOCK?.trim();
+    return lock || undefined;
+}