@cydm/pie 2.1.0 → 2.1.1

package/dist/builtin/skills/skill-creator/SKILL.md

@@ -382,12 +382,15 @@ Save the eval set to the workspace, then run in the background:
 node <skill-creator-path>/scripts/run_loop.mjs \
   --eval-set <path-to-trigger-eval.json> \
   --skill-path <path-to-skill> \
-  --model <model-id-powering-this-session> \
+  --model <profile/modelId> \
   --max-iterations 5 \
   --verbose
 ```
 
-Use the model ID from your system prompt (the one powering the current session) so the triggering test matches what the user actually experiences.
+Use a configured Pie `profile/modelId` so the triggering test matches what the
+user actually experiences. The runner applies selection through JSON-events
+metadata and verifies the actual session model. Unknown models and runtime
+errors fail explicitly; they are not counted as successful negative examples.
 
 While it runs, periodically tail the output to give the user updates on which iteration it's on and what the scores look like.
 

package/dist/builtin/skills/skill-creator/eval-viewer/generate_review.mjs

@@ -4,15 +4,14 @@ import path from "node:path";
 import { createServer } from "node:http";
 import { readFile, readdir, stat, writeFile } from "node:fs/promises";
 import { fileURLToPath } from "node:url";
-import { execFile } from "node:child_process";
-import { promisify } from "node:util";
 import { openInBrowser } from "../scripts/pie_runner.mjs";
+import { readJsonFile } from "../scripts/skill_metadata.mjs";
 
-const execFileAsync = promisify(execFile);
 const METADATA_FILES = new Set(["transcript.md", "user_notes.md", "metrics.json"]);
 const TEXT_EXTENSIONS = new Set([".txt", ".md", ".json", ".csv", ".py", ".js", ".ts", ".tsx", ".jsx", ".yaml", ".yml", ".xml", ".html", ".css", ".sh", ".rb", ".go", ".rs", ".java", ".c", ".cpp", ".h", ".hpp", ".sql", ".r", ".toml", ".mjs"]);
 const IMAGE_EXTENSIONS = new Set([".png", ".jpg", ".jpeg", ".gif", ".svg", ".webp"]);
 const MIME_OVERRIDES = {
+  ".pdf": "application/pdf",
   ".svg": "image/svg+xml",
   ".xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
   ".docx": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
@@ -36,7 +35,7 @@ async function pathExists(targetPath) {
 async function readJsonIfExists(targetPath) {
   if (!(await pathExists(targetPath))) return null;
   try {
-    return JSON.parse(await readFile(targetPath, "utf8"));
+    return await readJsonFile(targetPath);
   } catch {
     return null;
   }
@@ -175,18 +174,7 @@ export async function generateHtml({ workspace, skillName, previous = null, benc
   if (benchmark) {
     embedded.benchmark = benchmark;
   }
-  return template.replace("/*__EMBEDDED_DATA__*/", `const EMBEDDED_DATA = ${JSON.stringify(embedded)};`);
-}
-
-async function killPort(port) {
-  try {
-    const { stdout } = await execFileAsync("lsof", ["-ti", `:${port}`], { timeout: 5000 });
-    for (const pid of stdout.split("\n").map((value) => value.trim()).filter(Boolean)) {
-      try {
-        process.kill(Number(pid), "SIGTERM");
-      } catch {}
-    }
-  } catch {}
+  return template.replace("/*__EMBEDDED_DATA__*/", () => `const EMBEDDED_DATA = ${JSON.stringify(embedded).replaceAll("<", "\\u003c")};`);
 }
 
 function parseArgs(args) {
@@ -211,10 +199,10 @@ function parseArgs(args) {
 
 async function main(argv = process.argv) {
   const args = parseArgs(argv.slice(2));
-  const workspace = path.resolve(args._[0] ?? "");
-  if (!workspace) {
+  if (!args._[0]) {
     throw new Error("Usage: node generate_review.mjs <workspace-path> [--port PORT] [--skill-name NAME] [--static output.html]");
   }
+  const workspace = path.resolve(args._[0]);
   const skillName = args["skill-name"] ?? path.basename(workspace).replace(/-workspace$/, "");
   const previous = args["previous-workspace"] ? await loadPreviousIteration(path.resolve(args["previous-workspace"])) : {};
   const benchmark = args.benchmark ? await readJsonIfExists(path.resolve(args.benchmark)) : null;
@@ -225,7 +213,6 @@ async function main(argv = process.argv) {
     return;
   }
   let port = Number(args.port ?? args.p ?? 3117);
-  await killPort(port);
   const feedbackPath = path.join(workspace, "feedback.json");
   const server = createServer(async (req, res) => {
     try {
@@ -273,7 +260,7 @@ async function main(argv = process.argv) {
   port = typeof address === "object" && address ? address.port : port;
   const url = `http://localhost:${port}`;
   process.stdout.write(`Eval Viewer\nURL: ${url}\nWorkspace: ${workspace}\nFeedback: ${feedbackPath}\n`);
-  await openInBrowser(url);
+  if (!args["no-open"]) await openInBrowser(url);
 }
 
 const entryPath = fileURLToPath(import.meta.url);

package/dist/builtin/skills/skill-creator/eval-viewer/viewer.html

@@ -882,7 +882,7 @@
       const badgeClass = summary.pass_rate >= 0.8 ? "grade-pass" : summary.pass_rate >= 0.5 ? "" : "grade-fail";
       html += '<div class="grades-summary">';
       html += '<span class="grade-badge ' + badgeClass + '">' + passRate + '</span>';
-      html += '<span>' + (summary.passed || 0) + ' passed, ' + (summary.failed || 0) + ' failed of ' + (summary.total || 0) + '</span>';
+      html += '<span>' + escapeHtml(summary.passed || 0) + ' passed, ' + escapeHtml(summary.failed || 0) + ' failed of ' + escapeHtml(summary.total || 0) + '</span>';
       html += '</div>';
 
       // Assertions list
@@ -1099,7 +1099,7 @@
     function escapeHtml(text) {
       const div = document.createElement("div");
       div.textContent = text;
-      return div.innerHTML;
+      return div.innerHTML.replaceAll('"', "&quot;").replaceAll("'", "&#39;");
     }
 
     // ---- View switching ----
@@ -1129,16 +1129,16 @@
       html += "<h2 style='font-family: Poppins, sans-serif; margin-bottom: 0.5rem;'>Benchmark Results</h2>";
       html += "<p style='color: var(--text-muted); font-size: 0.875rem; margin-bottom: 1.25rem;'>";
       if (metadata.skill_name) html += "<strong>" + escapeHtml(metadata.skill_name) + "</strong> &mdash; ";
-      if (metadata.timestamp) html += metadata.timestamp + " &mdash; ";
-      if (metadata.evals_run) html += "Evals: " + metadata.evals_run.join(", ") + " &mdash; ";
-      html += (metadata.runs_per_configuration || "?") + " runs per configuration";
+      if (metadata.timestamp) html += escapeHtml(metadata.timestamp) + " &mdash; ";
+      if (metadata.evals_run) html += "Evals: " + escapeHtml(metadata.evals_run.join(", ")) + " &mdash; ";
+      html += escapeHtml(metadata.runs_per_configuration || "?") + " runs per configuration";
       html += "</p>";
 
       // Summary table
       html += '<table class="benchmark-table">';
 
       function fmtStat(stat, pct) {
-        if (!stat) return "—";
+        if (!stat || !Number.isFinite(stat.mean) || !Number.isFinite(stat.stddev)) return "—";
         const suffix = pct ? "%" : "";
         const m = pct ? (stat.mean * 100).toFixed(0) : stat.mean.toFixed(1);
         const s = pct ? (stat.stddev * 100).toFixed(0) : stat.stddev.toFixed(1);
@@ -1169,14 +1169,14 @@
       html += "<tr><td><strong>Pass Rate</strong></td>";
       html += "<td>" + fmtStat(a.pass_rate, true) + "</td>";
       html += "<td>" + fmtStat(b.pass_rate, true) + "</td>";
-      html += '<td class="' + deltaClass(delta.pass_rate) + '">' + (delta.pass_rate || "—") + "</td></tr>";
+      html += '<td class="' + deltaClass(delta.pass_rate) + '">' + escapeHtml(delta.pass_rate || "—") + "</td></tr>";
 
       // Time (only show row if data exists)
       if (a.time_seconds || b.time_seconds) {
         html += "<tr><td><strong>Time (s)</strong></td>";
         html += "<td>" + fmtStat(a.time_seconds, false) + "</td>";
         html += "<td>" + fmtStat(b.time_seconds, false) + "</td>";
-        html += '<td class="' + deltaClass(delta.time_seconds) + '">' + (delta.time_seconds ? delta.time_seconds + "s" : "—") + "</td></tr>";
+        html += '<td class="' + deltaClass(delta.time_seconds) + '">' + escapeHtml(delta.time_seconds ? delta.time_seconds + "s" : "—") + "</td></tr>";
       }
 
       // Tokens (only show row if data exists)
@@ -1184,7 +1184,7 @@
         html += "<tr><td><strong>Tokens</strong></td>";
         html += "<td>" + fmtStat(a.tokens, false) + "</td>";
         html += "<td>" + fmtStat(b.tokens, false) + "</td>";
-        html += '<td class="' + deltaClass(delta.tokens) + '">' + (delta.tokens || "—") + "</td></tr>";
+        html += '<td class="' + deltaClass(delta.tokens) + '">' + escapeHtml(delta.tokens || "—") + "</td></tr>";
       }
 
       html += "</tbody></table>";
@@ -1225,11 +1225,11 @@
               const r = run.result || {};
               const prClass = r.pass_rate >= 0.8 ? "benchmark-delta-positive" : r.pass_rate < 0.5 ? "benchmark-delta-negative" : "";
               html += '<tr class="' + rowClass + '">';
-              html += "<td>" + configLabel + "</td>";
-              html += "<td>" + run.run_number + "</td>";
-              html += '<td class="' + prClass + '">' + ((r.pass_rate || 0) * 100).toFixed(0) + "% (" + (r.passed || 0) + "/" + (r.total || 0) + ")</td>";
+              html += "<td>" + escapeHtml(configLabel) + "</td>";
+              html += "<td>" + escapeHtml(run.run_number) + "</td>";
+              html += '<td class="' + prClass + '">' + ((r.pass_rate || 0) * 100).toFixed(0) + "% (" + escapeHtml(r.passed || 0) + "/" + escapeHtml(r.total || 0) + ")</td>";
               if (hasTime) html += "<td>" + (r.time_seconds != null ? r.time_seconds.toFixed(1) : "—") + "</td>";
-              if (hasErrors) html += "<td>" + (r.errors || 0) + "</td>";
+              if (hasErrors) html += "<td>" + escapeHtml(r.errors || 0) + "</td>";
               html += "</tr>";
             }
 
@@ -1238,7 +1238,7 @@
             const avgRate = rates.reduce((a, b) => a + b, 0) / rates.length;
             const avgPrClass = avgRate >= 0.8 ? "benchmark-delta-positive" : avgRate < 0.5 ? "benchmark-delta-negative" : "";
             html += '<tr class="benchmark-row-avg ' + rowClass + '">';
-            html += "<td>" + configLabel + "</td>";
+            html += "<td>" + escapeHtml(configLabel) + "</td>";
             html += "<td>Avg</td>";
             html += '<td class="' + avgPrClass + '">' + (avgRate * 100).toFixed(0) + "%</td>";
             if (hasTime) {
@@ -1289,7 +1289,7 @@
                   if (exp) {
                     const cls = exp.passed ? "benchmark-delta-positive" : "benchmark-delta-negative";
                     const icon = exp.passed ? "\u2713" : "\u2717";
-                    html += '<span class="' + cls + '" title="Run ' + run.run_number + ': ' + escapeHtml(exp.evidence || "") + '">' + icon + "</span> ";
+                    html += '<span class="' + cls + '" title="Run ' + escapeHtml(run.run_number) + ': ' + escapeHtml(exp.evidence || "") + '">' + icon + "</span> ";
                   } else {
                     html += "— ";
                   }

package/dist/builtin/skills/skill-creator/scripts/aggregate_benchmark.mjs

@@ -1,8 +1,9 @@
 #!/usr/bin/env node
 
 import path from "node:path";
-import { readdir, readFile, stat, writeFile } from "node:fs/promises";
+import { readdir, stat, writeFile } from "node:fs/promises";
 import { fileURLToPath } from "node:url";
+import { readJsonFile } from "./skill_metadata.mjs";
 
 function round(value, places = 4) {
   const factor = 10 ** places;
@@ -41,7 +42,7 @@ async function readJsonIfExists(targetPath) {
     return null;
   }
   try {
-    return JSON.parse(await readFile(targetPath, "utf8"));
+    return await readJsonFile(targetPath);
   } catch {
     return null;
   }
@@ -76,10 +77,8 @@ export async function loadRunResults(benchmarkDir) {
     for (const configName of await listDirectories(evalDir)) {
       const configDir = path.join(evalDir, configName);
       const childDirs = await listDirectories(configDir);
-      const runDirs = childDirs.filter((name) => name.startsWith("run-"));
-      if (runDirs.length === 0) {
-        continue;
-      }
+      const nestedRuns = childDirs.filter((name) => name.startsWith("run-"));
+      const runDirs = nestedRuns.length > 0 ? nestedRuns : [""];
       if (!results[configName]) {
         results[configName] = [];
       }
@@ -91,19 +90,16 @@ export async function loadRunResults(benchmarkDir) {
         }
         const timing = grading.timing ?? {};
         let timeSeconds = Number(timing.total_duration_seconds ?? 0);
-        let tokens = 0;
+        let tokens = null;
         const timingJson = await readJsonIfExists(path.join(runDir, "timing.json"));
         if (timeSeconds === 0 && timingJson) {
           timeSeconds = Number(timingJson.total_duration_seconds ?? 0);
-          tokens = Number(timingJson.total_tokens ?? 0);
         }
+        if (Number.isFinite(timingJson?.total_tokens)) tokens = timingJson.total_tokens;
         const executionMetrics = grading.execution_metrics ?? {};
-        if (tokens === 0) {
-          tokens = Number(executionMetrics.output_chars ?? 0);
-        }
         results[configName].push({
           eval_id: evalId,
-          run_number: Number.parseInt(runDirName.split("-")[1], 10) || 0,
+          run_number: Number.parseInt(runDirName.split("-")[1], 10) || 1,
           pass_rate: Number(grading.summary?.pass_rate ?? 0),
           passed: Number(grading.summary?.passed ?? 0),
           failed: Number(grading.summary?.failed ?? 0),
@@ -127,13 +123,15 @@ export async function loadRunResults(benchmarkDir) {
 
 export function aggregateResults(results) {
   const runSummary = {};
-  const configs = Object.keys(results);
+  const configs = Object.keys(results).sort((a, b) => Number(b === "with_skill") - Number(a === "with_skill"));
   for (const config of configs) {
     const runs = results[config] ?? [];
     runSummary[config] = {
       pass_rate: calculateStats(runs.map((run) => Number(run.pass_rate ?? 0))),
       time_seconds: calculateStats(runs.map((run) => Number(run.time_seconds ?? 0))),
-      tokens: calculateStats(runs.map((run) => Number(run.tokens ?? 0))),
+      tokens: runs.some((run) => Number.isFinite(run.tokens))
+        ? calculateStats(runs.map((run) => run.tokens).filter(Number.isFinite))
+        : { mean: null, stddev: null, min: null, max: null },
     };
   }
   const primaryConfig = configs[0];
@@ -143,7 +141,8 @@ export function aggregateResults(results) {
   runSummary.delta = {
     pass_rate: `${(primary.pass_rate?.mean ?? 0) - (baseline.pass_rate?.mean ?? 0) >= 0 ? "+" : ""}${round((primary.pass_rate?.mean ?? 0) - (baseline.pass_rate?.mean ?? 0), 2).toFixed(2)}`,
     time_seconds: `${(primary.time_seconds?.mean ?? 0) - (baseline.time_seconds?.mean ?? 0) >= 0 ? "+" : ""}${round((primary.time_seconds?.mean ?? 0) - (baseline.time_seconds?.mean ?? 0), 1).toFixed(1)}`,
-    tokens: `${(primary.tokens?.mean ?? 0) - (baseline.tokens?.mean ?? 0) >= 0 ? "+" : ""}${Math.round((primary.tokens?.mean ?? 0) - (baseline.tokens?.mean ?? 0))}`,
+    tokens: primary.tokens?.mean != null && baseline.tokens?.mean != null
+      ? `${primary.tokens.mean - baseline.tokens.mean >= 0 ? "+" : ""}${Math.round(primary.tokens.mean - baseline.tokens.mean)}` : "—",
   };
   return runSummary;
 }
@@ -152,7 +151,8 @@ export async function generateBenchmark(benchmarkDir, { skillName = "", skillPat
   const results = await loadRunResults(benchmarkDir);
   const runSummary = aggregateResults(results);
   const runs = [];
-  for (const [configuration, entries] of Object.entries(results)) {
+  for (const configuration of Object.keys(runSummary).filter((key) => key !== "delta")) {
+    const entries = results[configuration];
     for (const result of entries) {
       runs.push({
         eval_id: result.eval_id,
@@ -174,6 +174,7 @@ export async function generateBenchmark(benchmarkDir, { skillName = "", skillPat
     }
   }
   const evalIds = [...new Set(runs.map((run) => run.eval_id))].sort((a, b) => a - b);
+  const counts = new Set(Object.values(results).flatMap((entries) => evalIds.map((id) => entries.filter((run) => run.eval_id === id).length)));
   return {
     metadata: {
       skill_name: skillName || "<skill-name>",
@@ -182,7 +183,7 @@ export async function generateBenchmark(benchmarkDir, { skillName = "", skillPat
       analyzer_model: "<model-name>",
       timestamp: new Date().toISOString().replace(/\.\d{3}Z$/, "Z"),
       evals_run: evalIds,
-      runs_per_configuration: 3,
+      runs_per_configuration: counts.size === 1 ? [...counts][0] : null,
     },
     runs,
     run_summary: runSummary,
@@ -206,7 +207,7 @@ export function generateMarkdown(benchmark) {
     "",
     `**Model**: ${metadata.executor_model ?? "<model-name>"}`,
     `**Date**: ${metadata.timestamp ?? ""}`,
-    `**Evals**: ${(metadata.evals_run ?? []).join(", ")} (${metadata.runs_per_configuration ?? 0} runs each per configuration)`,
+    `**Evals**: ${(metadata.evals_run ?? []).join(", ")} (${metadata.runs_per_configuration ?? "variable"} runs each per configuration)`,
     "",
     "## Summary",
     "",
@@ -214,7 +215,7 @@ export function generateMarkdown(benchmark) {
     "|--------|------------|---------------|-------|",
     `| Pass Rate | ${(aSummary.pass_rate?.mean ?? 0) * 100}% ± ${(aSummary.pass_rate?.stddev ?? 0) * 100}% | ${(bSummary.pass_rate?.mean ?? 0) * 100}% ± ${(bSummary.pass_rate?.stddev ?? 0) * 100}% | ${delta.pass_rate ?? "—"} |`,
     `| Time | ${(aSummary.time_seconds?.mean ?? 0).toFixed?.(1) ?? aSummary.time_seconds?.mean ?? 0}s ± ${(aSummary.time_seconds?.stddev ?? 0).toFixed?.(1) ?? aSummary.time_seconds?.stddev ?? 0}s | ${(bSummary.time_seconds?.mean ?? 0).toFixed?.(1) ?? bSummary.time_seconds?.mean ?? 0}s ± ${(bSummary.time_seconds?.stddev ?? 0).toFixed?.(1) ?? bSummary.time_seconds?.stddev ?? 0}s | ${delta.time_seconds ?? "—"}s |`,
-    `| Tokens | ${Math.round(aSummary.tokens?.mean ?? 0)} ± ${Math.round(aSummary.tokens?.stddev ?? 0)} | ${Math.round(bSummary.tokens?.mean ?? 0)} ± ${Math.round(bSummary.tokens?.stddev ?? 0)} | ${delta.tokens ?? "—"} |`,
+    `| Tokens | ${aSummary.tokens?.mean == null ? "—" : `${Math.round(aSummary.tokens.mean)} ± ${Math.round(aSummary.tokens.stddev)}`} | ${bSummary.tokens?.mean == null ? "—" : `${Math.round(bSummary.tokens.mean)} ± ${Math.round(bSummary.tokens.stddev)}`} | ${delta.tokens ?? "—"} |`,
   ];
   if (Array.isArray(benchmark.notes) && benchmark.notes.length > 0) {
     lines.push("", "## Notes", "");

package/dist/builtin/skills/skill-creator/scripts/generate_report.mjs

@@ -1,8 +1,9 @@
 #!/usr/bin/env node
 
 import path from "node:path";
-import { readFile, writeFile } from "node:fs/promises";
+import { writeFile } from "node:fs/promises";
 import { fileURLToPath } from "node:url";
+import { readJsonFile } from "./skill_metadata.mjs";
 
 function escapeHtml(value) {
   return String(value ?? "")
@@ -176,8 +177,7 @@ async function main(argv = process.argv) {
   if (!inputPath) {
     throw new Error("Usage: node generate_report.mjs <input.json> [-o output.html] [--skill-name name]");
   }
-  const raw = inputPath === "-" ? await readStdin() : await readFile(inputPath, "utf8");
-  const data = JSON.parse(raw);
+  const data = inputPath === "-" ? JSON.parse((await readStdin()).replace(/^\uFEFF/, "")) : await readJsonFile(inputPath);
   const output = generateHtml(data, { skillName: args["skill-name"] ?? "", autoRefresh: false });
   if (args.output || args.o) {
     await writeFile(args.output || args.o, output);

package/dist/builtin/skills/skill-creator/scripts/improve_description.mjs

@@ -1,10 +1,10 @@
 #!/usr/bin/env node
 
 import path from "node:path";
-import { mkdir, readFile, writeFile } from "node:fs/promises";
+import { mkdir, writeFile } from "node:fs/promises";
 import { fileURLToPath } from "node:url";
 import { callPieText } from "./pie_runner.mjs";
-import { parseSkillFile } from "./skill_metadata.mjs";
+import { parseSkillFile, readJsonFile, validateFrontmatter } from "./skill_metadata.mjs";
 
 export async function improveDescription({
   skillName,
@@ -100,8 +100,11 @@ I'd encourage you to be creative and mix up the style in different iterations si
 
 Please respond with only the new description text in <new_description> tags, nothing else.`;
 
-  const transcript = { iteration, prompt };
-  const text = await callPieText(prompt);
+  const transcript = { iteration, prompt, requested_model: model, actual_models: [] };
+  const onModelSelection = ({ actualModel }) => {
+    if (actualModel && !transcript.actual_models.includes(actualModel)) transcript.actual_models.push(actualModel);
+  };
+  const text = await callPieText(prompt, { model, onModelSelection });
   transcript.response = text;
 
   let description = parseDescription(text);
@@ -119,7 +122,7 @@ A previous attempt produced this description, which at ${description.length} cha
 "${description}"
 
 Rewrite it to be under 1024 characters while keeping the most important trigger words and intent coverage. Respond with only the new description in <new_description> tags.`;
-    const shortenText = await callPieText(shortenPrompt);
+    const shortenText = await callPieText(shortenPrompt, { model, onModelSelection });
     const shortened = parseDescription(shortenText);
     transcript.rewrite_prompt = shortenPrompt;
     transcript.rewrite_response = shortenText;
@@ -133,7 +136,8 @@ Rewrite it to be under 1024 characters while keeping the most important trigger
     await mkdir(logDir, { recursive: true });
     await writeFile(path.join(logDir, `improve_iter_${iteration ?? "unknown"}.json`), JSON.stringify(transcript, null, 2));
   }
-
+  const validation = validateFrontmatter({ name: skillName, description });
+  if (!validation.valid) throw new Error(`Improved description is invalid: ${validation.message}`);
   return description;
 }
 
@@ -145,8 +149,8 @@ function parseDescription(text) {
 async function main(argv = process.argv) {
   const args = parseArgs(argv.slice(2));
   const skillPath = path.resolve(args["skill-path"]);
-  const evalResults = JSON.parse(await readFile(args["eval-results"], "utf8"));
-  const history = args.history ? JSON.parse(await readFile(args.history, "utf8")) : [];
+  const evalResults = await readJsonFile(args["eval-results"]);
+  const history = args.history ? await readJsonFile(args.history) : [];
   const { name, content } = await parseSkillFile(skillPath);
   const currentDescription = evalResults.description;
 

package/dist/builtin/skills/skill-creator/scripts/package_skill.mjs

@@ -81,6 +81,7 @@ export async function packageSkill(skillPath, outputDir) {
   const files = await collectFiles(resolvedSkillPath, baseDir);
 
   for (const file of files) {
+    if (path.resolve(file.fullPath) === archivePath) continue;
     if (shouldExclude(file.arcname)) {
       process.stdout.write(`  Skipped: ${file.arcname}\n`);
       continue;

package/dist/builtin/skills/skill-creator/scripts/pie_runner.mjs

@@ -3,6 +3,7 @@
 import crypto from "node:crypto";
 import { spawn } from "node:child_process";
 import fs from "node:fs";
+import os from "node:os";
 import path from "node:path";
 import { fileURLToPath } from "node:url";
 
@@ -61,21 +62,29 @@ export function getPieCliPath() {
   );
 }
 
-export async function runPieJsonPrompt(prompt, { cwd = process.cwd(), timeout = 300_000, sessionId = null } = {}) {
+export async function runPieJsonPrompt(prompt, { cwd = process.cwd(), timeout = 300_000, sessionId = null, model = null } = {}) {
+  if (model && (typeof model !== "string" || !/^[^/\s]+\/\S+$/.test(model))) {
+    throw new Error("Model must use profile/modelId format.");
+  }
   return new Promise((resolve, reject) => {
-    const args = [getPieCliPath(), prompt, "--json-output"];
-    if (sessionId) {
-      args.push("--session-id", sessionId);
-    }
+    const id = sessionId || `skill-run-${uuidFragment()}`;
+    const turnId = `skill-turn-${uuidFragment()}`;
+    const args = [getPieCliPath(), "--json-events", "--event-detail", "full", "--session-id", id];
 
-    const child = spawn("node", args, {
+    const child = spawn(process.execPath, args, {
       cwd,
       env: { ...process.env },
-      stdio: ["ignore", "pipe", "pipe"],
+      stdio: ["pipe", "pipe", "pipe"],
+      windowsHide: true,
     });
+    child.stdout.setEncoding("utf8");
+    child.stderr.setEncoding("utf8");
 
-    let stdout = "";
+    let buffer = "";
     let stderr = "";
+    let text = "";
+    let response;
+    let turnSettled = false;
     let settled = false;
 
     const cleanup = () => {
@@ -89,55 +98,92 @@ export async function runPieJsonPrompt(prompt, { cwd = process.cwd(), timeout =
       if (settled) return;
       settled = true;
       cleanup();
-      try {
-        child.kill("SIGKILL");
-      } catch {}
       cb();
     };
 
-    const tryResolveFromStdout = () => {
-      const lines = stdout.split(/\r?\n/).map((line) => line.trim()).filter(Boolean);
-      for (let i = lines.length - 1; i >= 0; i -= 1) {
-        const line = lines[i];
-        if (!line.startsWith("{")) continue;
-        try {
-          const parsed = JSON.parse(line);
-          if (Object.prototype.hasOwnProperty.call(parsed, "ok")) {
-            finish(() => resolve(parsed));
-            return true;
-          }
-        } catch {}
+    const terminate = () => {
+      child.kill();
+      child.stdin.destroy();
+      child.stdout.destroy();
+      child.stderr.destroy();
+      child.unref();
+    };
+
+    const consume = (line) => {
+      let event;
+      try { event = JSON.parse(line); } catch { return; }
+      if (event.turnId && event.turnId !== turnId) return;
+      if (event.type === "display.output" && event.stream === "terminal") text += event.chunk || "";
+      if (event.type === "turn.completed") response = { ok: true, text: event.summary ?? text };
+      if (event.type === "turn.failed" || event.type === "turn.cancelled") {
+        const error = event.error || event.reason || event.type;
+        response = { ok: false, error: typeof error === "string" ? error : error.message || JSON.stringify(error) };
       }
-      return false;
+      if (event.type === "turn.settled") turnSettled = true;
     };
 
     const timer = setTimeout(() => {
+      terminate();
       finish(() => reject(new Error(`pie prompt timed out after ${Math.round(timeout / 1000)}s`)));
     }, timeout);
 
     child.stdout.on("data", (chunk) => {
-      stdout += String(chunk);
-      tryResolveFromStdout();
+      buffer += String(chunk);
+      let newline;
+      while ((newline = buffer.indexOf("\n")) >= 0) {
+        consume(buffer.slice(0, newline).trim());
+        buffer = buffer.slice(newline + 1);
+      }
     });
     child.stderr.on("data", (chunk) => {
-      stderr += String(chunk);
+      stderr = (stderr + String(chunk)).slice(-8_000);
     });
     child.on("error", (error) => {
       finish(() => reject(error));
     });
     child.on("close", (code) => {
-      if (tryResolveFromStdout()) {
+      if (buffer.trim()) consume(buffer.trim());
+      if (code !== 0 || !turnSettled || !response) {
+        finish(() => reject(new Error(`pie prompt exited ${code} without a settled response\nstderr: ${stderr}`)));
+        return;
+      }
+      const configuredDir = process.env.PIE_CODING_AGENT_DIR || path.join(os.homedir(), ".pie");
+      const configDir = configuredDir === "~" ? os.homedir()
+        : configuredDir.startsWith("~/") ? path.join(os.homedir(), configuredDir.slice(2)) : configuredDir;
+      let actualModel = null;
+      try {
+        const session = JSON.parse(fs.readFileSync(path.join(configDir, "sessions", `${id}.json`), "utf8"));
+        const metadata = session.metadata || {};
+        if (metadata.provider && metadata.modelId) actualModel = `${metadata.provider}/${metadata.modelId}`;
+        const messages = session.messages || (session.entries || []).map((entry) => entry.message).filter(Boolean);
+        const lastAssistant = messages.filter((message) => message.role === "assistant").at(-1);
+        if (response.ok && lastAssistant) {
+          response.text = typeof lastAssistant.content === "string" ? lastAssistant.content
+            : (lastAssistant.content || []).filter((block) => block.type === "text").map((block) => block.text).join("");
+        }
+      } catch {}
+      if (model && response.ok && actualModel !== model) {
+        finish(() => reject(new Error(`Requested model ${model}, but session reported ${actualModel || "no model"}.`)));
         return;
       }
-      finish(() => reject(new Error(`pie prompt exited ${code}\nstdout: ${stdout}\nstderr: ${stderr}`)));
+      finish(() => resolve({ ...response, sessionId: id, requestedModel: model, actualModel }));
     });
+    child.stdin.end(`${JSON.stringify({
+      type: "turn.request",
+      sessionId: id,
+      turnId,
+      prompt,
+      cwd,
+      ...(model ? { metadata: { model } } : {}),
+    })}\n`);
   });
 }
 
-export async function callPieText(prompt, { cwd = process.cwd(), timeout = 300_000, sessionId = null } = {}) {
-  const response = await runPieJsonPrompt(prompt, { cwd, timeout, sessionId });
+export async function callPieText(prompt, { onModelSelection, ...options } = {}) {
+  const response = await runPieJsonPrompt(prompt, options);
   if (!response?.ok) {
     throw new Error(response?.error || "Pie prompt failed");
   }
+  onModelSelection?.({ requestedModel: response.requestedModel, actualModel: response.actualModel });
   return String(response.text || "");
 }