github · nicolehaugen · Sep 17, 2026 · Sep 17, 2026 · Sep 17, 2026 · Sep 17, 2026
@@ -5,7 +5,7 @@
   },
   "metadata": {
     "description": "Spec Kit integrations for GitHub Copilot CLI and the GitHub Copilot App.",
-    "version": "0.18.1"
+    "version": "0.19.5"
   },
   "plugins": [
     {
@@ -34,8 +34,8 @@
     },
     {
       "name": "spec-kit-copilot-wizard",
-      "description": "Adds a guided Spec Kit Wizard canvas that drives the full spec-driven development lifecycle via the spec-kit-copilot skills plugin.",
-      "version": "0.1.1",
+      "description": "Adds a guided Spec Kit Wizard canvas that drives composed pipelines and generates project canvases from selected phases.",
+      "version": "0.2.5",
       "source": "plugins/spec-kit-copilot-wizard"
     }
   ]

@@ -2,10 +2,12 @@
 node_modules/
 .playwright-mcp/
 
+# Local Impeccable hook cache
+.impeccable/hook.cache.json
+
 # Session artifacts (canvas runtime state, spec-kit init output, css coverage snapshots)
 .speckit-wizard/
 .specify/
 .github/skills/
 css-baseline-*.png
 css-c*.png
-
@@ -43,7 +43,7 @@ Contributions are welcome — see [CONTRIBUTING.md](CONTRIBUTING.md) to get star
 | `spec-kit-copilot-assess` | 0.1.0 | Copilot App canvas | Optional visual dashboard for the Spec Kit `assess` extension |
 | `spec-kit-copilot-bugfix` | 0.1.0 | Copilot App canvas | Optional visual dashboard for the Spec Kit `bug` extension |
 | `spec-kit-copilot-sdd` | 0.1.0 | Copilot App canvas | Optional visual dashboard for the core spec-driven development workflow |
-| `spec-kit-copilot-wizard` | 0.1.1 | Copilot App canvas | Optional guided wizard canvas for the full Spec Kit lifecycle |
+| `spec-kit-copilot-wizard` | 0.2.5 | Copilot App canvas | Guided lifecycle composer that can generate project canvases from selected phases |
 
 The plugins are independently installable and versioned. Install the core skills,
 the assessment canvas, the bug fix canvas, the spec-driven development canvas, the
@@ -100,7 +100,7 @@ own README for full details.
 | [`assess-canvas`](plugins/spec-kit-copilot-assess/extensions/assess-canvas/README.md) | `spec-kit-copilot-assess` | Dashboard for the optional `assess` extension — the intake → research → define → shape → decide funnel. |
 | [`bugfix-canvas`](plugins/spec-kit-copilot-bugfix/extensions/bugfix-canvas/README.md) | `spec-kit-copilot-bugfix` | Dashboard for the optional `bug` extension — the assess → fix → test triage pipeline. |
 | [`sdd-canvas`](plugins/spec-kit-copilot-sdd/extensions/sdd-canvas/README.md) | `spec-kit-copilot-sdd` | Dashboard for the core spec-driven workflow — constitution → specify → clarify → plan → tasks → analyze → checklist → implement. |
-| [`speckit-wizard-canvas`](plugins/spec-kit-copilot-wizard/extensions/speckit-wizard-canvas/README.md) | `spec-kit-copilot-wizard` | Guided wizard for the full Spec Kit lifecycle — setup → constitution → specify → clarify → plan → tasks → analyze → checklist → implement, with preset / extension / composition inspectors. |
+| [`speckit-wizard-canvas`](plugins/spec-kit-copilot-wizard/extensions/speckit-wizard-canvas/README.md) | `spec-kit-copilot-wizard` | Guided lifecycle composer with preset / extension / composition inspectors and one-click generation of project canvases from the selected phase pipeline. |
 
 ### Previews
 

@@ -42,7 +42,8 @@ import {
 // instead of waiting for the agent turn to finish. Agent-side errors still
 // surface in chat; transport/session failures are observed asynchronously so
 // local tracking state can be cleaned up without blocking the caller.
-export function dispatchPromptToSession({ prompt, onError } = {}) {
+export function dispatchPromptToSession({ prompt, onError, waitForAcceptance = false } = {}) {
+    if (waitForAcceptance) return sessionAdapter().send({ prompt });
     setImmediate(() => {
         let completion;
         try {

@@ -55,6 +55,7 @@ export function newInstance(instanceId) {
         _stateWatchLastMtimeMs: 0, // last processed mtime to suppress echoes
         artifactWatchers: [],    // fs.watch handles on .specify / specs dirs
         _artifactWatchDebounce: null, // pending debounce timer for artifact rescans
+        generation: null,         // latest generated-canvas request/result
     };
 }
 

@@ -11,6 +11,7 @@
 // server does not emit HTML fragments.
 
 import { PHASE_ORDER, PHASE_BY_ID } from "./wizard-phases.mjs";
+import { CORE_CAPABILITIES } from "../pipeline/canonical.mjs";
 import { deriveSetupPhaseStatus } from "../state/store.mjs";
 
 // -------- Compact state snapshot ------------------------------------------
@@ -33,6 +34,7 @@ export function buildStateSnapshot(scan) {
             environment: null,
             boot: null,
             depsError: null,
+            generation: null,
             warnings: [],
         };
     }
@@ -54,6 +56,12 @@ export function buildStateSnapshot(scan) {
             formValues: scanned.formValues ?? {},
             artifactPath: scanned.artifactPath ?? null,
         };
+        if (id === "checklist") {
+            const folder = scanned.artifactPath?.endsWith("/") ? scanned.artifactPath : scanned.folderPath;
+            phases[id].artifactTemplatePath = scanned.artifactTemplatePath ?? CORE_CAPABILITIES["speckit.checklist"].writesTo;
+            if (folder) phases[id].folderPath = folder.replace(/<slug>/g, scan.slug ?? "<slug>").replace(/\/$/, "");
+            if (scanned.artifactPath?.endsWith("/")) phases[id].artifactPath = null;
+        }
     }
     // Forward any extension-command status slices set by the scanner (keyed
     // `commands/<full-cmd-id>`). PHASE_ORDER only enumerates core phases, so
@@ -68,6 +76,7 @@ export function buildStateSnapshot(scan) {
                 id,
                 status: slice?.status ?? "empty",
                 artifactPath: slice?.artifactPath ?? null,
+                artifactTemplatePath: slice?.artifactTemplatePath ?? null,
                 lastRunAt: slice?.lastRunAt ?? null,
                 formValues: slice?.formValues ?? {},
                 // LLM-inferred metadata from artifact-targets.json cache (via
@@ -146,6 +155,7 @@ export function buildStateSnapshot(scan) {
         environment: scan.environment ?? null,
         boot: scan.boot ?? null,
         depsError: scan.depsError ?? null,
+        generation: scan.generation ?? null,
         scaffoldedSkills: Array.isArray(scan.scaffoldedSkills) ? scan.scaffoldedSkills : [],
         warnings: Array.isArray(scan.warnings) ? scan.warnings.slice(0, 20) : [],
     };
@@ -178,8 +188,9 @@ function buildCommands(scan, statusPhases) {
         // phase (constitution, specify, plan, tasks, analyze, checklist).
         // Otherwise default to "empty" — the runtime interaction loop will
         // mark it done via /api/phase/status when Copilot writes the artifact.
-        const status = statusPhases?.[cmd.id]?.status ?? "empty";
-        let artifactPath = statusPhases?.[cmd.id]?.artifactPath ?? cmd.artifact ?? null;
+        const phaseSlice = commandPhaseSlice(statusPhases, cmd);
+        const status = phaseSlice?.status ?? "empty";
+        let artifactPath = phaseSlice?.artifactPath ?? cmd.artifact ?? null;
         if (typeof artifactPath === "string" && artifactPath.includes("<slug>") && scan.slug) {
             artifactPath = artifactPath.replace(/<slug>/g, scan.slug);
         }
@@ -197,12 +208,19 @@ function buildCommands(scan, statusPhases) {
             id: cmd.id,
             commandName: cmd.name,
             shortLabel: deriveShortLabel(cmd.name, cmd.id),
-            title: cmd.description || cmd.name,
-            helpText: cmd.description || "",
+            title: phaseSlice?.description || cmd.description || cmd.name,
+            helpText: phaseSlice?.description || cmd.description || "",
             handoffs,
             optional: !!cmd.optional,
             artifact: cmd.artifact ?? null,
             artifactPath,
+            ...(phaseSlice?.artifactTemplatePath ? {
+                artifactTemplatePath: phaseSlice.id === "checklist"
+                    ? cmd.artifact ?? phaseSlice.artifactTemplatePath : phaseSlice.artifactTemplatePath,
+            } : {}),
+            ...(phaseSlice?.folderPath ? { folderPath: phaseSlice.folderPath } : {}),
+            ...(phaseSlice?.argsHint ? { argsHint: phaseSlice.argsHint } : {}),
+            ...(phaseSlice?.argsWhenEmpty ? { argsWhenEmpty: phaseSlice.argsWhenEmpty } : {}),
             status,
             locked: !setupGateOpen,
             source: cmd.source ?? "core",
@@ -211,6 +229,21 @@ function buildCommands(scan, statusPhases) {
     return out;
 }
 
+function commandPhaseSlice(statusPhases, command) {
+    if (!statusPhases || typeof statusPhases !== "object") return null;
+    const candidates = [command?.id, command?.name];
+    for (const candidate of candidates) {
+        if (typeof candidate !== "string" || !candidate) continue;
+        const bare = candidate.startsWith("commands/")
+            ? candidate.slice("commands/".length)
+            : candidate;
+        for (const key of [candidate, bare, `commands/${bare}`]) {
+            if (statusPhases[key]) return statusPhases[key];
+        }
+    }
+    return null;
+}
+
 /**
  * Derive a short, human-friendly stepper label from a command name / id.
  * Examples: "speckit.constitution" -> "Constitution",

@@ -55,6 +55,7 @@ import { scanWorkspace } from "../project-scanner.mjs";
 import { buildStateSnapshot } from "./snapshot-builder.mjs";
 import { applyPatch, overlayCachedComposition, activeFingerprint } from "../state/store.mjs";
 import { fsDeps } from "./instances.mjs";
+import { recoverGenerationStatus } from "../generation/storage.mjs";
 
 export async function snapshot(inst) {
     // Preset precedence: consume the order the `speckit-preset` skill
@@ -159,6 +160,10 @@ export async function snapshot(inst) {
     // /api/skills/reload) so the UI can gate setup completion on the
     // live SDK result rather than a persisted flag or a folder probe.
     snap.skillsReload = inst.skillsReload ?? null;
+    // Generated-canvas requests are durable. Re-read the latest request/result
+    // pair so extension reloads and fresh SSE subscriptions recover progress.
+    inst.generation = await recoverGenerationStatus(inst.workspacePath).catch(() => inst.generation ?? null);
+    snap.generation = inst.generation ?? null;
     inst.state = applyPatch(inst.state ?? {}, {
         currentPhase: scan.currentPhase,
         preset: scan.preset,

@@ -19,39 +19,48 @@ import { hydrateFromCatalogSources, cliOrderFromInstalled, specifyRun } from "./
 // still functional, just missing the "added" badge.
 export async function listInstalledExtensions(workspacePath) {
     const stdout = await specifyRun(["extension", "list"], workspacePath);
+    return parseExtensionListOutput(stdout);
+}
+
+export function parseExtensionListOutput(stdout) {
     const ids = new Set();
     const names = new Set();
+    const byId = new Map();
     const byName = new Map();
     const orderedIds = [];
-    if (stdout == null) return { ids, names, byName, orderedIds };
+    if (typeof stdout !== "string") return { ids, names, byId, byName, orderedIds };
     // `specify extension list` prints two-line entries:
     //     ✓ <Display Name> (v<version>)
     //        <extension-id>
     //        <description...>
     // We parse the header + following non-empty line as the id.
     const lines = stdout.split(/\r?\n/);
     for (let i = 0; i < lines.length; i++) {
-        const header = lines[i].match(/^\s*[✓✗x]\s+(.+?)\s+\(v[^)]+\)\s*$/);
+        const header = lines[i].match(/^\s*([✓✗x])\s+(.+?)\s+\(v[^)]+\)\s*$/);
         if (!header) continue;
-        const name = header[1].trim();
+        const enabled = header[1] === "✓";
+        const name = header[2].trim();
         // Find the next non-empty line — that's the id.
         let id = null;
+        let priority = null;
         for (let j = i + 1; j < lines.length; j++) {
             const t = lines[j].trim();
             if (!t) continue;
             // Stop if we've reached the next header row.
             if (/^[✓✗x]\s+.+\(v[^)]+\)\s*$/.test(t)) break;
-            id = t.split(/\s+/)[0];
-            break;
+            if (!id) id = t.split(/\s+/)[0];
+            const priorityMatch = t.match(/\bpriority\s*:?\s*(\d+)\b/i);
+            if (priorityMatch) priority = Number(priorityMatch[1]);
         }
         if (id) {
             names.add(name.toLowerCase());
             ids.add(id);
+            byId.set(id, { id, name, enabled, priority });
             byName.set(name.toLowerCase(), id);
             orderedIds.push(id);
         }
     }
-    return { ids, names, byName, orderedIds };
+    return { ids, names, byId, byName, orderedIds };
 }
 
 // Given extension catalog sources, fetch each source's JSON directly to
@@ -65,6 +74,8 @@ export async function hydrateExtensionsForSources(inst, sources) {
         outputField: "cachedExtensionItems",
         listInstalled: listInstalledExtensions,
         extraFields: (_raw, { installedId, installed }) => ({
+            enabled: installedId ? installed.byId?.get(installedId)?.enabled ?? null : null,
+            priority: installedId ? installed.byId?.get(installedId)?.priority ?? null : null,
             // CLI precedence position from `specify extension list` (0 = first
             // line = winner). null when the extension isn't installed. See
             // the same field on preset items for the rationale — the wizard

@@ -25,11 +25,12 @@ import { hydrateFromCatalogSources, cliOrderFromInstalled, specifyRun } from "./
 export async function listInstalledPresets(workspacePath) {
     const stdout = await specifyRun(["preset", "list"], workspacePath);
     if (stdout == null) {
-        return { ids: new Set(), names: new Set(), byName: new Map(), orderedIds: [] };
+        return { ids: new Set(), names: new Set(), byId: new Map(), byName: new Map(), orderedIds: [] };
     }
     const parsed = parsePresetListOutput(stdout);
     return {
         ids: new Set(parsed.orderedIds),
+        byId: parsed.byId,
         names: new Set(parsed.byName.keys()),
         byName: parsed.byName,
         // CLI precedence order (first = winner). Consumed by the
@@ -50,6 +51,8 @@ export async function hydratePresetsForSources(inst, sources) {
         outputField: "cachedPresetItems",
         listInstalled: listInstalledPresets,
         extraFields: (_raw, { installedId, installed }) => ({
+            enabled: installedId ? installed.byId?.get(installedId)?.enabled ?? null : null,
+            priority: installedId ? installed.byId?.get(installedId)?.priority ?? null : null,
             // CLI precedence position (0 = first line of `specify preset list`
             // = winner). null when the preset isn't installed, or when the CLI
             // list wasn't available. The assembler uses this as the primary

@@ -16,7 +16,7 @@ import { fetchCatalogJson } from "./sources.mjs";
 import { spawn } from "node:child_process";
 import { buildAugmentedPath } from "../env/resolve-path.mjs";
 
-const EMPTY_INSTALLED = Object.freeze({ ids: new Set(), names: new Set(), byName: new Map(), orderedIds: [] });
+const EMPTY_INSTALLED = Object.freeze({ ids: new Set(), names: new Set(), byId: new Map(), byName: new Map(), orderedIds: [] });
 
 // Memoize the augmented PATH lookup. This runs on every `specify` invocation
 // (list installed, etc.), so scanning SDK/uv/pipx dirs once per process is

@@ -58,7 +58,7 @@
  * @param {string} stdout
  * @returns {{
  *   orderedIds: string[],
- *   byId: Map<string, { id: string, name: string, version: string|null, enabled: boolean }>,
+ *   byId: Map<string, { id: string, name: string, version: string|null, enabled: boolean, priority: number|null }>,
  *   byName: Map<string, string>,
  * }}
  */
@@ -75,7 +75,9 @@ export function parsePresetListOutput(stdout) {
     // The name may itself contain parentheses (e.g. "Some Preset (Full)"),
     // so we match the LAST "(<id>) v<version>" pair on the line, then treat
     // everything before it as the display name.
-    for (const raw of stdout.split(/\r?\n/)) {
+    const lines = stdout.split(/\r?\n/);
+    for (let index = 0; index < lines.length; index++) {
+        const raw = lines[index];
         const m = raw.match(/^\s+(.+?)\s+\(([^()]+)\)\s+v([\d.]+)(?:\s+[—-]\s+(enabled|disabled))?/i);
         if (!m) continue;
         const name = m[1].trim();
@@ -87,14 +89,28 @@ export function parsePresetListOutput(stdout) {
         // (which only lists installed presets and prints "disabled" only
         // when explicitly disabled).
         const enabled = enabledToken === "" || enabledToken === "enabled";
+        let priority = priorityFromLine(raw);
+        if (priority === null) {
+            for (let cursor = index + 1; cursor < lines.length; cursor++) {
+                const next = lines[cursor];
+                if (/^\s+.+?\s+\([^()]+\)\s+v[\d.]+/i.test(next)) break;
+                priority = priorityFromLine(`${raw} ${next.trim()}`);
+                if (priority !== null) break;
+            }
+        }
         if (byId.has(id)) continue; // defensive against duplicate parses
         orderedIds.push(id);
-        byId.set(id, { id, name, version, enabled });
+        byId.set(id, { id, name, version, enabled, priority });
         byName.set(name.toLowerCase(), id);
     }
     return { orderedIds, byId, byName };
 }
 
+function priorityFromLine(line) {
+    const match = String(line ?? "").match(/\bpriority\s*:?\s*(\d+)\b/i);
+    return match ? Number(match[1]) : null;
+}
+
 /**
  * Reorder an array of loaded presets to match the skill-declared order.
  * Any preset not present in `orderedIds` is appended at the end in the

@@ -0,0 +1,51 @@
+// Determine whether a composed pipeline can be materialized as a standalone canvas.
+import { dirname } from "node:path/posix";
+
+function itemRoot(pathTemplate) {
+    if (typeof pathTemplate !== "string" || !pathTemplate.includes("<slug>")) return null;
+    const marker = pathTemplate.indexOf("<slug>");
+    return pathTemplate.slice(0, marker + "<slug>".length);
+}
+
+export function assessVisualizationApplicability(steps, workflowSteps = steps) {
+    const errors = [];
+    const itemRoots = new Set();
+
+    for (const step of steps) {
+        const expected = step.index === 0 ? [] : [step.index - 1];
+        if (JSON.stringify(step.predecessors) !== JSON.stringify(expected)) {
+            errors.push({
+                code: "visualization_unsupported",
+                path: `pipeline[${step.index}].predecessors`,
+                message: `Phase "${step.label}" is not part of a simple linear predecessor chain.`,
+            });
+        }
+        const artifact = step.artifact?.pathTemplate;
+        if (artifact && !artifact.toLowerCase().endsWith(".md")) {
+            errors.push({
+                code: "visualization_unsupported",
+                path: `pipeline[${step.index}].artifact`,
+                message: `Phase "${step.label}" requires a non-Markdown artifact viewer.`,
+            });
+        }
+        const root = workflowSteps.includes(step) ? itemRoot(artifact) : null;
+        if (root) itemRoots.add(root);
+    }
+
+    if (itemRoots.size > 1) {
+        errors.push({
+            code: "visualization_unsupported",
+            path: "pipeline",
+            message: "The pipeline uses multiple independent item roots that the standard workflow canvas cannot represent.",
+        });
+    }
+
+    const root = [...itemRoots][0] ?? null;
+    return {
+        ok: errors.length === 0,
+        errors,
+        workflowMode: root ? "item" : "project",
+        itemRoot: root,
+        itemDirectory: root ? dirname(root) : null,
+    };
+}