gsd-pi 2.67.0-dev.a5b1d8f → 2.67.0-dev.fe39184

package/README.md

@@ -27,56 +27,66 @@ One command. Walk away. Come back to a built project with clean git history.
 
 ---
 
-## What's New in v2.63.0
+## What's New in v2.67
 
-### MCP Server & Integrations
+### Context Engineering
+
+- **Tiered Context Injection (M005)** — relevance-scoped context with 65%+ token reduction. Decision scope cascade derives context from slice metadata instead of blanket injection.
+- **Resilient transient error recovery** — defers to Core RetryHandler and fixes cmdCtx race conditions for more reliable auto-mode sessions.
 
-- **MCP server** — 6 read-only project state tools for external integrations, auto-wrapup guard, and question dedup.
-- **Ollama extension** — first-class local LLM support via Ollama, with dynamic routing enabled by default.
-- **Discord bot & daemon** — dedicated daemon package, Discord bot with 6 discord.js shard event listeners, and headless text mode with tool calls.
+### Provider & Model Improvements
 
-### Intelligent Model Routing
+- **Anthropic subscription routing** — users with Anthropic subscriptions are automatically routed through Claude Code CLI provider with proper display names across all UI surfaces.
+- **Claude Code provider hardening** — native Windows claude lookup, fallback guards, and `out of extra usage` error matching.
+- **XML parameter recovery** — pi-ai recovers XML parameters trapped in JSON strings from providers.
 
-- **Capability-aware model routing (ADR-004)** — capability scoring, `before_model_select` hook, and task metadata extraction replace pattern-based model selection.
-- **Stop/backtrack capture classifications** — context optimization with model routing and masking.
+### Safety & Data Integrity
 
-### VS Code & TUI
+- **LLM safety harness** — auto-mode damage control prevents the LLM from running destructive operations or querying `gsd.db` directly via bash.
+- **5-wave state machine hardening** — critical data integrity fixes across atomic writes, randomized tmp paths, event log reconciliation, session recovery, and consistency enforcement. 86+ regression tests added.
+- **Discussion gate enforcement** — mechanical enforcement for discussion question gates with fail-closed behavior.
+- **Enhanced verification** — pre-execution plan verification checks, post-execution cross-task consistency checks, blocking behavior and strict mode.
 
-- **VS Code sidebar redesign** — SCM provider, checkpoints, diagnostics panel, activity feed, workflow controls, session forking, and enhanced code lens.
-- **`/gsd parallel watch`** — native TUI overlay for real-time worker monitoring.
-- **Real-time TUI monitor dashboard** — self-healing dashboard with colorized headless verbose output.
+### Parallel Execution & Dispatch
 
-### Agent & Auto-Mode
+- **Slice-level parallelism** — dependency-aware parallel dispatch within a milestone, not just across milestones.
+- **Parallel research slices** — research and milestone validation run in parallel.
+- **Worker model override** — configure different models for parallel milestone workers.
 
-- **`/btw` skill** — ephemeral side questions without interrupting the main workflow.
-- **Codebase map** — automatic codebase map injection for fresh agent contexts.
-- **`--resume` flag** — resume previous sessions from the CLI.
-- **Concurrent invocation guard** — prevents overlapping auto-mode runs with parallel worker reliability improvements.
-- **Safety mechanisms on by default** — snapshots and pre-merge checks enabled out of the box.
+### TUI & Notifications
 
-### Infrastructure & Performance
+- **Persistent notification panel** — TUI overlay, widget, and web API for real-time notifications.
+- **Remote questions race** — local TUI races against remote channel (Slack/Discord) instead of remote-only routing.
+- **OS-specific keyboard shortcuts** — shortcut hints now adapt to macOS/Linux/Windows.
+- **`/gsd show-config`** — inspect active configuration at a glance.
 
-- **Topological sort for extensions** — correct load ordering for dependent extensions.
-- **Headless integration hardening** — real-time streaming, verbose output, and observability improvements across v2.54–v2.55.
-- **GLM-5.1 model** — new model support added.
-- **80+ bug fixes** — worktree safety, parallel mode scoping, state corruption, and more.
+### Infrastructure
+
+- **Ollama native provider** — `/api/chat` provider with full option exposure, `apiKey` auth mode, and headless probe.
+- **MCP OAuth** — MCP client supports OAuth auth provider for HTTP transport.
+- **WAL-safe migration backup** — database migrations create WAL-safe backups with stronger regression tests.
+- **Xcode/xcodegen detection** — project detection now supports Xcode bundles and xcodegen.
+- **170+ bug fixes** — state machine resilience, worktree safety, prompt injection, session recovery, and more.
 
 See the full [Changelog](./CHANGELOG.md) for details on every release.
 
 <details>
-<summary>Previous highlights (v2.52 and earlier)</summary>
-
+<summary>Previous highlights (v2.63 and earlier)</summary>
+
+- **MCP server** — 6 read-only project state tools for external integrations, auto-wrapup guard, and question dedup
+- **Ollama extension** — first-class local LLM support via Ollama, with dynamic routing enabled by default
+- **Discord bot & daemon** — dedicated daemon package, Discord bot, and headless text mode with tool calls
+- **Capability-aware model routing (ADR-004)** — capability scoring, `before_model_select` hook, and task metadata extraction
+- **VS Code sidebar redesign** — SCM provider, checkpoints, diagnostics panel, activity feed, workflow controls, session forking
+- **`/gsd parallel watch`** — native TUI overlay for real-time worker monitoring
+- **Codebase map** — automatic codebase map injection for fresh agent contexts
+- **`--resume` flag** — resume previous sessions from the CLI
+- **Concurrent invocation guard** — prevents overlapping auto-mode runs
 - **VS Code integration** — status bar, file decorations, bash terminal, session tree, conversation history, and code lens
-- **Capability-based model selection** — replaced model-ID pattern matching with capability metadata
 - **Skills overhaul** — 30+ skill packs covering major frameworks, databases, and cloud platforms
-- **Quality gates** — 8-question quality gates for planning and completion templates
 - **Single-writer state engine** — disciplined state transitions with machine guards and TOCTOU hardening
-- **`/gsd rethink`** — conversational project reorganization
-- **Complete offline mode** — fully offline with local models
-- **Browser-based web interface** — run GSD from the browser with `gsd --web`
 - **DB-backed planning tools** — atomic SQLite tool calls for state transitions
 - **Declarative workflow engine** — YAML workflows through auto-loop
-- **7 data-loss prevention fixes** — hallucination guard, merge anchor verification, dirty tree detection, and more
 - **Doctor: worktree lifecycle checks** — validates worktree health, detects orphans, consolidates cleanup
 
 </details>

package/dist/resources/extensions/claude-code-cli/stream-adapter.js

@@ -9,6 +9,7 @@
 import { EventStream } from "@gsd/pi-ai";
 import { execSync } from "node:child_process";
 import { PartialMessageBuilder, ZERO_USAGE, mapUsage } from "./partial-builder.js";
+import { buildWorkflowMcpServers } from "../gsd/workflow-mcp.js";
 // ---------------------------------------------------------------------------
 // Stream factory
 // ---------------------------------------------------------------------------
@@ -133,6 +134,7 @@ export function makeStreamExhaustedErrorMessage(model, lastTextContent) {
  * beta flags, and other configuration without mocking the full SDK.
  */
 export function buildSdkOptions(modelId, prompt) {
+    const mcpServers = buildWorkflowMcpServers();
     return {
         pathToClaudeCodeExecutable: getClaudePath(),
         model: modelId,
@@ -143,9 +145,98 @@ export function buildSdkOptions(modelId, prompt) {
         allowDangerouslySkipPermissions: true,
         settingSources: ["project"],
         systemPrompt: { type: "preset", preset: "claude_code" },
+        ...(mcpServers ? { mcpServers } : {}),
         betas: modelId.includes("sonnet") ? ["context-1m-2025-08-07"] : [],
     };
 }
+function normalizeToolResultContent(content) {
+    if (typeof content === "string") {
+        return [{ type: "text", text: content }];
+    }
+    if (!Array.isArray(content)) {
+        if (content == null)
+            return [{ type: "text", text: "" }];
+        return [{ type: "text", text: JSON.stringify(content) }];
+    }
+    const blocks = [];
+    for (const item of content) {
+        if (typeof item === "string") {
+            blocks.push({ type: "text", text: item });
+            continue;
+        }
+        if (!item || typeof item !== "object") {
+            blocks.push({ type: "text", text: String(item) });
+            continue;
+        }
+        const block = item;
+        if (block.type === "text") {
+            blocks.push({ type: "text", text: typeof block.text === "string" ? block.text : "" });
+            continue;
+        }
+        if (block.type === "image"
+            && typeof block.data === "string"
+            && typeof block.mimeType === "string") {
+            blocks.push({ type: "image", data: block.data, mimeType: block.mimeType });
+            continue;
+        }
+        blocks.push({ type: "text", text: JSON.stringify(block) });
+    }
+    return blocks.length > 0 ? blocks : [{ type: "text", text: "" }];
+}
+export function extractToolResultsFromSdkUserMessage(message) {
+    const extracted = [];
+    const seen = new Set();
+    const rawMessage = message.message;
+    const content = Array.isArray(rawMessage?.content) ? rawMessage.content : [];
+    for (const item of content) {
+        if (!item || typeof item !== "object")
+            continue;
+        const block = item;
+        const type = typeof block.type === "string" ? block.type : "";
+        if (type !== "tool_result" && type !== "mcp_tool_result")
+            continue;
+        const toolUseId = typeof block.tool_use_id === "string" ? block.tool_use_id : "";
+        if (!toolUseId || seen.has(toolUseId))
+            continue;
+        seen.add(toolUseId);
+        extracted.push({
+            toolUseId,
+            result: {
+                content: normalizeToolResultContent(block.content),
+                details: {},
+                isError: block.is_error === true,
+            },
+        });
+    }
+    if (extracted.length === 0) {
+        const fallback = message.tool_use_result;
+        if (fallback && typeof fallback === "object") {
+            const toolResult = fallback;
+            const toolUseId = typeof toolResult.tool_use_id === "string" ? toolResult.tool_use_id : "";
+            if (toolUseId) {
+                extracted.push({
+                    toolUseId,
+                    result: {
+                        content: normalizeToolResultContent(toolResult.content),
+                        details: {},
+                        isError: toolResult.is_error === true,
+                    },
+                });
+            }
+        }
+    }
+    return extracted;
+}
+function attachExternalResultsToToolCalls(toolCalls, toolResultsById) {
+    for (const block of toolCalls) {
+        if (block.type !== "toolCall")
+            continue;
+        const externalResult = toolResultsById.get(block.id);
+        if (!externalResult)
+            continue;
+        block.externalResult = externalResult;
+    }
+}
 // ---------------------------------------------------------------------------
 // streamSimple implementation
 // ---------------------------------------------------------------------------
@@ -169,6 +260,8 @@ async function pumpSdkMessages(model, context, options, stream) {
     let lastThinkingContent = "";
     /** Collect tool calls from intermediate SDK turns for tool_execution events. */
     const intermediateToolCalls = [];
+    /** Preserve real external tool results from Claude Code's synthetic user messages. */
+    const toolResultsById = new Map();
     try {
         // Dynamic import — the SDK is an optional dependency.
         const sdkModule = "@anthropic-ai/claude-agent-sdk";
@@ -221,14 +314,7 @@ async function pumpSdkMessages(model, context, options, stream) {
                         break;
                     const assistantEvent = builder.handleEvent(event);
                     if (assistantEvent) {
-                        // Skip toolcall events — the agent loop's externalToolExecution
-                        // path emits tool_execution_start/end events after streamSimple
-                        // returns. Streaming toolcall events would render tool calls
-                        // out of order in the TUI's accumulated message content.
-                        const t = assistantEvent.type;
-                        if (t !== "toolcall_start" && t !== "toolcall_delta" && t !== "toolcall_end") {
-                            stream.push(assistantEvent);
-                        }
+                        stream.push(assistantEvent);
                     }
                     break;
                 }
@@ -263,6 +349,32 @@ async function pumpSdkMessages(model, context, options, stream) {
                             }
                         }
                     }
+                    // Extract tool results from the SDK's synthetic user message
+                    // and attach to corresponding tool call blocks immediately.
+                    for (const { toolUseId, result } of extractToolResultsFromSdkUserMessage(msg)) {
+                        toolResultsById.set(toolUseId, result);
+                    }
+                    attachExternalResultsToToolCalls(intermediateToolCalls, toolResultsById);
+                    // Push a synthetic toolcall_end for each tool call from this turn
+                    // so the TUI can render tool results in real-time during the SDK
+                    // session instead of waiting until the entire session completes.
+                    if (builder) {
+                        for (const block of builder.message.content) {
+                            if (block.type !== "toolCall")
+                                continue;
+                            const extResult = block.externalResult;
+                            if (!extResult)
+                                continue;
+                            // Push a toolcall_end with result attached so the chat-controller
+                            // can call updateResult on the pending ToolExecutionComponent.
+                            stream.push({
+                                type: "toolcall_end",
+                                contentIndex: builder.message.content.indexOf(block),
+                                toolCall: block,
+                                partial: builder.message,
+                            });
+                        }
+                    }
                     builder = null;
                     break;
                 }
@@ -274,6 +386,7 @@ async function pumpSdkMessages(model, context, options, stream) {
                     // events for proper TUI rendering, followed by the text response.
                     const finalContent = [];
                     // Add tool calls from intermediate turns first (renders above text)
+                    attachExternalResultsToToolCalls(intermediateToolCalls, toolResultsById);
                     finalContent.push(...intermediateToolCalls);
                     // Add text/thinking from the last turn
                     if (builder && builder.message.content.length > 0) {

package/dist/resources/extensions/gsd/auto/phases.js

@@ -27,6 +27,7 @@ import { isDbAvailable, getMilestoneSlices } from "../gsd-db.js";
 import { resetEvidence } from "../safety/evidence-collector.js";
 import { createCheckpoint, cleanupCheckpoint, rollbackToCheckpoint } from "../safety/git-checkpoint.js";
 import { resolveSafetyHarnessConfig } from "../safety/safety-harness.js";
+import { getWorkflowTransportSupportError, getRequiredWorkflowToolsForAutoUnit, } from "../workflow-mcp.js";
 // ─── generateMilestoneReport ──────────────────────────────────────────────────
 /**
  * Resolve the base path for milestone reports.
@@ -886,6 +887,22 @@ export async function runUnitPhase(ic, iterData, loopState, sidecarItem) {
     s.currentDispatchedModelId = s.currentUnitModel
         ? `${s.currentUnitModel.provider ?? ""}/${s.currentUnitModel.id ?? ""}`
         : null;
+    const compatibilityError = getWorkflowTransportSupportError(s.currentUnitModel?.provider ?? ctx.model?.provider, getRequiredWorkflowToolsForAutoUnit(unitType), {
+        projectRoot: s.basePath,
+        surface: "auto-mode",
+        unitType,
+        authMode: s.currentUnitModel?.provider
+            ? ctx.modelRegistry.getProviderAuthMode(s.currentUnitModel.provider)
+            : ctx.model?.provider
+                ? ctx.modelRegistry.getProviderAuthMode(ctx.model.provider)
+                : undefined,
+        baseUrl: s.currentUnitModel?.baseUrl ?? ctx.model?.baseUrl,
+    });
+    if (compatibilityError) {
+        ctx.ui.notify(compatibilityError, "error");
+        await deps.stopAuto(ctx, pi, compatibilityError);
+        return { action: "break", reason: "workflow-capability" };
+    }
     // Progress widget + preconditions — deferred to after model selection so the
     // widget's first render tick shows the correct model (#2899).
     deps.updateProgressWidget(ctx, unitType, unitId, state);

package/dist/resources/extensions/gsd/auto/session.js

@@ -33,6 +33,9 @@ export class AutoSession {
     // ── Paths ────────────────────────────────────────────────────────────────
     basePath = "";
     originalBasePath = "";
+    previousProjectRootEnv = null;
+    hadProjectRootEnv = false;
+    projectRootEnvCaptured = false;
     gitService = null;
     // ── Dispatch counters ────────────────────────────────────────────────────
     unitDispatchCount = new Map();
@@ -132,6 +135,9 @@ export class AutoSession {
         // Paths
         this.basePath = "";
         this.originalBasePath = "";
+        this.previousProjectRootEnv = null;
+        this.hadProjectRootEnv = false;
+        this.projectRootEnvCaptured = false;
         this.gitService = null;
         // Dispatch
         this.unitDispatchCount.clear();

package/dist/resources/extensions/gsd/auto-direct-dispatch.js

@@ -10,6 +10,7 @@ import { resolveMilestoneFile, resolveSliceFile, relSliceFile, } from "./paths.j
 import { buildResearchSlicePrompt, buildResearchMilestonePrompt, buildPlanSlicePrompt, buildPlanMilestonePrompt, buildExecuteTaskPrompt, buildCompleteSlicePrompt, buildCompleteMilestonePrompt, buildReassessRoadmapPrompt, buildRunUatPrompt, buildReplanSlicePrompt, } from "./auto-prompts.js";
 import { loadEffectiveGSDPreferences } from "./preferences.js";
 import { pauseAuto } from "./auto.js";
+import { getWorkflowTransportSupportError, getRequiredWorkflowToolsForAutoUnit, } from "./workflow-mcp.js";
 export async function dispatchDirectPhase(ctx, pi, phase, base) {
     const state = await deriveState(base);
     const mid = state.activeMilestone?.id;
@@ -202,6 +203,17 @@ export async function dispatchDirectPhase(ctx, pi, phase, base) {
             ctx.ui.notify(`Unknown phase "${phase}". Valid phases: research, plan, execute, complete, reassess, uat, replan.`, "warning");
             return;
     }
+    const compatibilityError = getWorkflowTransportSupportError(ctx.model?.provider, getRequiredWorkflowToolsForAutoUnit(unitType), {
+        projectRoot: base,
+        surface: "direct phase dispatch",
+        unitType,
+        authMode: ctx.model?.provider ? ctx.modelRegistry.getProviderAuthMode(ctx.model.provider) : undefined,
+        baseUrl: ctx.model?.baseUrl,
+    });
+    if (compatibilityError) {
+        ctx.ui.notify(compatibilityError, "error");
+        return;
+    }
     ctx.ui.notify(`Dispatching ${unitType} for ${unitId}...`, "info");
     const result = await ctx.newSession();
     if (result.cancelled) {

package/dist/resources/extensions/gsd/auto-start.js

@@ -250,6 +250,18 @@ export async function bootstrapAutoSession(s, ctx, pi, base, verboseMode, reques
                 logWarning("engine", `mkdir failed: ${err instanceof Error ? err.message : String(err)}`);
             }
         }
+        if (ctx.model?.provider === "claude-code") {
+            try {
+                const { ensureProjectWorkflowMcpConfig } = await import("./mcp-project-config.js");
+                const result = ensureProjectWorkflowMcpConfig(base);
+                if (result.status !== "unchanged") {
+                    ctx.ui.notify(`Claude Code MCP prepared at ${result.configPath}`, "info");
+                }
+            }
+            catch (err) {
+                ctx.ui.notify(`Claude Code MCP prep failed: ${err instanceof Error ? err.message : String(err)}`, "warning");
+            }
+        }
         // Initialize GitServiceImpl
         s.gitService = new GitServiceImpl(s.basePath, loadEffectiveGSDPreferences()?.preferences?.git ?? {});
         // Check for crash from previous session. Skip our own fresh bootstrap lock.

package/dist/resources/extensions/gsd/auto.js

@@ -79,6 +79,27 @@ export { MAX_UNIT_DISPATCHES, STUB_RECOVERY_THRESHOLD, MAX_LIFETIME_DISPATCHES,
 const s = new AutoSession();
 /** Throttle STATE.md rebuilds — at most once per 30 seconds */
 const STATE_REBUILD_MIN_INTERVAL_MS = 30_000;
+function captureProjectRootEnv(projectRoot) {
+    if (!s.projectRootEnvCaptured) {
+        s.hadProjectRootEnv = Object.prototype.hasOwnProperty.call(process.env, "GSD_PROJECT_ROOT");
+        s.previousProjectRootEnv = process.env.GSD_PROJECT_ROOT ?? null;
+        s.projectRootEnvCaptured = true;
+    }
+    process.env.GSD_PROJECT_ROOT = projectRoot;
+}
+function restoreProjectRootEnv() {
+    if (!s.projectRootEnvCaptured)
+        return;
+    if (s.hadProjectRootEnv && s.previousProjectRootEnv !== null) {
+        process.env.GSD_PROJECT_ROOT = s.previousProjectRootEnv;
+    }
+    else {
+        delete process.env.GSD_PROJECT_ROOT;
+    }
+    s.previousProjectRootEnv = null;
+    s.hadProjectRootEnv = false;
+    s.projectRootEnvCaptured = false;
+}
 export function shouldUseWorktreeIsolation() {
     const prefs = loadEffectiveGSDPreferences()?.preferences?.git;
     if (prefs?.isolation === "worktree")
@@ -315,6 +336,7 @@ function handleLostSessionLock(ctx, lockStatus) {
     s.active = false;
     s.paused = false;
     clearUnitTimeout();
+    restoreProjectRootEnv();
     deregisterSigtermHandler();
     clearCmuxSidebar(loadEffectiveGSDPreferences()?.preferences);
     const base = lockBase();
@@ -345,6 +367,7 @@ function cleanupAfterLoopExit(ctx) {
     s.currentUnit = null;
     s.active = false;
     clearUnitTimeout();
+    restoreProjectRootEnv();
     // Clear crash lock and release session lock so the next `/gsd next` does
     // not see a stale lock with the current PID and treat it as a "remote"
     // session (which would cause it to SIGTERM itself). (#2730)
@@ -601,6 +624,7 @@ export async function stopAuto(ctx, pi, reason) {
         ctx?.ui.setStatus("gsd-auto", undefined);
         ctx?.ui.setWidget("gsd-progress", undefined);
         ctx?.ui.setFooter(undefined);
+        restoreProjectRootEnv();
         // Reset all session state in one call
         s.reset();
     }
@@ -674,6 +698,7 @@ export async function pauseAuto(ctx, _pi, _errorContext) {
     _resetPendingResolve();
     s.active = false;
     s.paused = true;
+    restoreProjectRootEnv();
     s.pendingVerificationRetry = null;
     s.verificationRetryCount.clear();
     ctx?.ui.setStatus("gsd-auto", "paused");
@@ -981,6 +1006,7 @@ export async function startAuto(ctx, pi, base, verboseMode, options) {
         updateSessionLock(lockBase(), "resuming", s.currentMilestoneId ?? "unknown");
         writeLock(lockBase(), "resuming", s.currentMilestoneId ?? "unknown");
         logCmuxEvent(loadEffectiveGSDPreferences()?.preferences, s.stepMode ? "Step-mode resumed." : "Auto-mode resumed.", "progress");
+        captureProjectRootEnv(s.originalBasePath || s.basePath);
         await autoLoop(ctx, pi, s, buildLoopDeps());
         cleanupAfterLoopExit(ctx);
         return;
@@ -995,6 +1021,7 @@ export async function startAuto(ctx, pi, base, verboseMode, options) {
     const ready = await bootstrapAutoSession(s, ctx, pi, base, verboseMode, requestedStepMode, bootstrapDeps);
     if (!ready)
         return;
+    captureProjectRootEnv(s.originalBasePath || s.basePath);
     try {
         syncCmuxSidebar(loadEffectiveGSDPreferences()?.preferences, await deriveState(s.basePath));
     }