Keep lab links inside the published static site

CHANGELOG.md

@@ -1,5 +1,10 @@
 # Changelog
 
+## 0.11.1 — 2026-09-15
+
+- Use explicit HTML filenames for published lab navigation. Hugging Face redirects bare directory paths to Hub routes rather than serving each directory index. Cover homepage entries and research/skill-lab return links.
+- Reseal only the published lab presentation files; recorded trials, raw JSON and original research archives stay byte-identical. Test navigation with a static server that rejects implicit directory indexes.
+
 ## 0.11.0 — 2026-09-15
 
 - Add offline AgentCore Evaluate import, golden-case/rubric comparison and recomputable trace review with responsive standalone reports.

README.md

@@ -117,11 +117,11 @@ also requires all checks to pass. [Verification and limits](docs/verification.md
 
 ## Trace Workbench — 0.11.0
 
-Import saved AgentCore Evaluate responses, versioned golden cases and Skills Anywhere delivery receipts. Inspect zero scores, skipped judges, missing results and missed skills separately. Compare matching datasets/rubrics and verify preserved input bytes offline. [Try the authored controls](https://noteflowai.github.io/evalarc/trace-workbench/) · [Actual local MCP delivery](https://noteflowai.github.io/evalarc/trace-mcp/) · [Input contract](docs/trace-workbench.md). No live AWS evaluation is claimed.
+Import saved AgentCore Evaluate responses, versioned golden cases and Skills Anywhere delivery receipts. Inspect zero scores, skipped judges, missing results and missed skills separately. Compare matching datasets/rubrics and verify preserved input bytes offline. [Try the authored controls](https://noteflowai.github.io/evalarc/trace-workbench/index.html) · [Actual local MCP delivery](https://noteflowai.github.io/evalarc/trace-mcp/index.html) · [Input contract](docs/trace-workbench.md). No live AWS evaluation is claimed.
 
 ## New in 0.9.0: research you can inspect
 
-[Explore all 27 real GPU skill trials](https://noteflowai.github.io/evalarc/skill-impact/) and [the research pilots](docs/research-pilots.md). Robot Reel's [captured-scene editor](https://noteflowai.github.io/robot-reel/scene-lab/) and [official LIBERO-Plus replay](https://noteflowai.github.io/robot-reel/libero-plus/) connect real source records with portable skill delivery and independent grading. Every failed attempt stays visible; no skill efficacy, full-benchmark or real-hardware result is implied.
+[Explore all 27 real GPU skill trials](https://noteflowai.github.io/evalarc/skill-impact/index.html) and [the research pilots](docs/research-pilots.md). Robot Reel's [captured-scene editor](https://noteflowai.github.io/robot-reel/scene-lab/) and [official LIBERO-Plus replay](https://noteflowai.github.io/robot-reel/libero-plus/) connect real source records with portable skill delivery and independent grading. Every failed attempt stays visible; no skill efficacy, full-benchmark or real-hardware result is implied.
 
 
 ## Run an audit

README.zh-CN.md

@@ -100,11 +100,11 @@ Docker 评测：参考策略 3/3 轮完全通过，重复写入策略虽然平
 
 ## 0.11.0：运行记录评估工作台
 
-导入已保存的 AgentCore Evaluate 结果、版本化黄金案例和 Skills Anywhere 加载回执，分别查看有效零分、评估跳过、缺少结果及漏调用技能。支持相同测试集与评分规则下的对比，以及原始输入的离线复核。[交互示例](https://noteflowai.github.io/evalarc/trace-workbench/) · [真实本地 MCP 加载](https://noteflowai.github.io/evalarc/trace-mcp/) · [数据契约](docs/trace-workbench.md)。示例明确区分合成评分与真实加载记录，未运行云端评估。
+导入已保存的 AgentCore Evaluate 结果、版本化黄金案例和 Skills Anywhere 加载回执，分别查看有效零分、评估跳过、缺少结果及漏调用技能。支持相同测试集与评分规则下的对比，以及原始输入的离线复核。[交互示例](https://noteflowai.github.io/evalarc/trace-workbench/index.html) · [真实本地 MCP 加载](https://noteflowai.github.io/evalarc/trace-mcp/index.html) · [数据契约](docs/trace-workbench.md)。示例明确区分合成评分与真实加载记录，未运行云端评估。
 
 ## 0.9.0：有原始证据的研究场景
 
-[查看 27 次真实 GPU 技能评测](https://noteflowai.github.io/evalarc/skill-impact/)，并阅读[完整方法与限制](docs/research-pilots.md)。新增[实景 Blender 编辑](https://noteflowai.github.io/robot-reel/scene-lab/)与[官方 LIBERO-Plus 子集回放](https://noteflowai.github.io/robot-reel/libero-plus/)，把原始记录、技能交付与独立验收连接起来。失败尝试全部保留；不宣称技能提分、完整基准成绩或真机效果。
+[查看 27 次真实 GPU 技能评测](https://noteflowai.github.io/evalarc/skill-impact/index.html)，并阅读[完整方法与限制](docs/research-pilots.md)。新增[实景 Blender 编辑](https://noteflowai.github.io/robot-reel/scene-lab/)与[官方 LIBERO-Plus 子集回放](https://noteflowai.github.io/robot-reel/libero-plus/)，把原始记录、技能交付与独立验收连接起来。失败尝试全部保留；不宣称技能提分、完整基准成绩或真机效果。
 
 
 ## 直接运行

package.json

@@ -1,6 +1,6 @@
 {
   "name": "evalarc-evidence-site",
-  "version": "0.11.0",
+  "version": "0.11.1",
   "private": true,
   "description": "Browser checks for the static EvalArc evidence explorer",
   "scripts": {

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "evalarc"
-version = "0.11.0"
+version = "0.11.1"
 description = "Auditable task environments and evaluations for coding and tool-using agents."
 readme = "README.md"
 requires-python = ">=3.11"

scripts/build_site.py

@@ -225,6 +225,30 @@ def sha256(path: Path) -> str:
     return hashlib.sha256(path.read_bytes()).hexdigest()
 
 
+def explicit_lab_navigation(folder: Path, commit: str) -> None:
+    """Update copied presentation links, preserving every original experiment byte."""
+    page = folder / "index.html"
+    original = page.read_bytes()
+    updated = original.replace(b'href="../"', b'href="../index.html"').replace(
+        b'href="../skill-impact/"', b'href="../skill-impact/index.html"'
+    )
+    if updated == original:
+        return
+    page.write_bytes(updated)
+    manifest_path = folder / "manifest.json"
+    manifest = json.loads(manifest_path.read_bytes())
+    manifest["presentation"] = {
+        "source_commit": commit,
+        "original_index_sha256": hashlib.sha256(original).hexdigest(),
+        "change": "Explicit index.html navigation for static hosting",
+    }
+    manifest["files"]["index.html"] = {
+        "sha256": hashlib.sha256(updated).hexdigest(),
+        "bytes": len(updated),
+    }
+    manifest_path.write_text(json.dumps(manifest, indent=2) + "\n")
+
+
 def verify(folder: Path) -> dict:
     record = json.loads((folder / MANIFEST).read_text())
     if record["source_repository"] != SOURCE or not re.fullmatch(
@@ -394,6 +418,10 @@ def build(destination: Path) -> dict:
     verify_records(ROOT / "examples/research")
     shutil.copytree(ROOT / "examples/skill-impact", destination / "skill-impact")
     shutil.copytree(ROOT / "examples/research", destination / "research")
+    for lab in ("skill-impact", "research"):
+        explicit_lab_navigation(destination / lab, commit)
+    verify_lab(destination / "skill-impact")
+    verify_records(destination / "research")
     from evalarc.trace_review import import_trace
 
     trace_examples = ROOT / "examples/trace-workbench"

scripts/check_site.cjs

@@ -13,7 +13,12 @@ async function main() {
     const mime = {".html":"text/html", ".js":"text/javascript", ".json":"application/json", ".css":"text/css", ".svg":"image/svg+xml", ".zip":"application/zip"};
     server = http.createServer((req, res) => {
       const route = decodeURIComponent(new URL(req.url, "http://localhost").pathname);
-      const filename = path.resolve(root, "." + (route.endsWith("/") ? route + "index.html" : route));
+      // HF static hosting does not resolve subdirectory indexes. Keep that
+      // production constraint in the local browser fixture.
+      if (route !== "/" && route.endsWith("/")) {
+        res.writeHead(404).end(); return;
+      }
+      const filename = path.resolve(root, "." + (route === "/" ? "/index.html" : route));
       if (!filename.startsWith(root + path.sep) || !fs.existsSync(filename) || !fs.statSync(filename).isFile()) {
         res.writeHead(404).end(); return;
       }
@@ -226,7 +231,7 @@ async function main() {
         await app.getByRole("heading", {name:"support-routing", exact:true}).waitFor();
         assert.match(await app.locator("body").innerText(), /retry-after-commit/);
       }
-      await app.locator("body").evaluate((element, url) => { location.href = new URL("skill-impact/", url).href; }, appUrl);
+      await app.locator("body").evaluate((element, url) => { location.href = new URL("skill-impact/index.html", url).href; }, appUrl);
       await app.locator("#workspace").waitFor({state:"visible"});
       assert.equal(await app.locator(".trial").count(), 9);
       assert.match(await app.locator("#profile-totals").innerText(), /2\/9 fully resolved/);
@@ -248,10 +253,14 @@ async function main() {
         fs.mkdirSync(process.env.RESEARCH_SCREENSHOTS,{recursive:true});
         await page.screenshot({path:path.join(process.env.RESEARCH_SCREENSHOTS,"skill-impact.png"),fullPage:true});
       }
-      await app.locator("body").evaluate((element, url) => { location.href = new URL("research/", url).href; }, appUrl);
+      await app.locator("body").evaluate((element, url) => { location.href = new URL("research/index.html", url).href; }, appUrl);
       await app.getByRole("heading",{name:"Skill composition: 12 attempts, 3 accepted"}).waitFor();
       assert.equal(await app.locator("tbody tr").count(),18);
       assert.equal(await app.locator("body").evaluate(() => document.documentElement.scrollWidth > innerWidth), false);
+      await app.getByRole("link",{name:"27-trial Skill Impact Lab"}).click();
+      await app.locator("#profile").waitFor();
+      await app.getByRole("link",{name:"Playground"}).click();
+      await app.locator("#workspace").waitFor({state:"visible"});
       const coveragePage = await browser.newPage({viewport:{width,height:1000}});
       try {
         await coveragePage.goto(new URL("robot/index.html#fault-3", suiteBase).href);
@@ -298,7 +307,7 @@ async function main() {
       assert.equal(await page.locator("body").evaluate(() => document.documentElement.scrollWidth > innerWidth), false);
       assert.deepEqual(errors, []);
       failures.add("skill-impact/lab.json");
-      await page.goto(new URL("skill-impact/",base).href);
+      await page.goto(new URL("skill-impact/index.html",base).href);
       await page.locator("#retry").waitFor({state:"visible"});
       failures.clear();
       await page.locator("#retry").click();
@@ -361,7 +370,8 @@ async function main() {
         const errors = [];
         page.on("pageerror", error => errors.push(error.message));
         try {
-          await page.goto(new URL("trace-workbench/",base).href);
+          await page.goto(base);
+          await page.getByRole("link",{name:"Explore five authored review controls"}).click();
           await page.locator("#filter-status").filter({hasText:"5 cases shown"}).waitFor();
           assert.equal(await page.locator("article[data-gate=accepted]").count(),1);
           assert.equal(await page.locator("article[data-gate=rejected]").count(),2);
@@ -387,7 +397,8 @@ async function main() {
           assert.equal(await download.failure(),null);
           const original = fs.readFileSync(path.join(root,"trace-workbench/input.json"));
           assert.deepEqual(fs.readFileSync(await download.path()),original);
-          await page.goto(new URL("trace-mcp/",base).href);
+          await page.goto(base);
+          await page.getByRole("link",{name:"Inspect an actual MCP delivery"}).click();
           await page.locator("#filter-status").filter({hasText:"1 cases shown"}).waitFor();
           assert.match(await page.locator("body").innerText(),/Actual local stdio MCP/);
           assert.match(await page.locator("article").innerText(),/MATCHED/);

site/index.html

@@ -37,8 +37,8 @@ <h1>Look past<br>the <em>score.</em></h1>
       </a>
     </section>
     <p class="muted">Generate Python or JavaScript candidates and audit independent controls against the same task contracts. <a href="https://github.com/noteflowai/evalarc/blob/main/docs/languages.md">Try both runtimes &#8599;</a>. The recorded showcases below retain their original versions and fingerprints.</p>
-    <section class="suite" aria-label="Real GPU research pilots"><p class="eyebrow">NEW / RECORDED MODEL EVIDENCE</p><h2>A skill loaded. Did the task pass?</h2><p>27 real Qwen3-8B trials compare no skill, direct loading and MCP delivery on attributed robot recordings. Inspect every tool receipt, candidate, independent score and ATIF trajectory. Three engineering profiles, including negative results.</p><div class="comparison-links"><a href="skill-impact/">Explore the Skill Impact Lab &#8599;</a><a href="research/">Composition, handoff &amp; Harbor evidence &#8599;</a></div><p class="caption">L40S recordings; one public development task. No skill accuracy gain or general model ranking is claimed.</p></section>
-    <section class="suite" aria-label="Trace Workbench"><p class="eyebrow">NEW / BRING YOUR AGENT RECORDS</p><h2>Zero, skipped, or missing?</h2><p>A zero can be a valid judgment. A skipped evaluator needs context. A required skill may never have loaded. Inspect each against a versioned golden case, with recording identity and explicit acceptance rules.</p><div class="comparison-links"><a href="trace-workbench/">Explore five authored review controls &#8599;</a><a href="trace-mcp/">Inspect an actual MCP delivery &#8599;</a></div><p class="caption">The five controls are synthetic. The separate MCP record contains a real local load and no evaluator scores. Review your own saved AgentCore Evaluate export offline with <code>evalarc trace-import</code>.</p></section>
+    <section class="suite" aria-label="Real GPU research pilots"><p class="eyebrow">NEW / RECORDED MODEL EVIDENCE</p><h2>A skill loaded. Did the task pass?</h2><p>27 real Qwen3-8B trials compare no skill, direct loading and MCP delivery on attributed robot recordings. Inspect every tool receipt, candidate, independent score and ATIF trajectory. Three engineering profiles, including negative results.</p><div class="comparison-links"><a href="skill-impact/index.html">Explore the Skill Impact Lab &#8599;</a><a href="research/index.html">Composition, handoff &amp; Harbor evidence &#8599;</a></div><p class="caption">L40S recordings; one public development task. No skill accuracy gain or general model ranking is claimed.</p></section>
+    <section class="suite" aria-label="Trace Workbench"><p class="eyebrow">NEW / BRING YOUR AGENT RECORDS</p><h2>Zero, skipped, or missing?</h2><p>A zero can be a valid judgment. A skipped evaluator needs context. A required skill may never have loaded. Inspect each against a versioned golden case, with recording identity and explicit acceptance rules.</p><div class="comparison-links"><a href="trace-workbench/index.html">Explore five authored review controls &#8599;</a><a href="trace-mcp/index.html">Inspect an actual MCP delivery &#8599;</a></div><p class="caption">The five controls are synthetic. The separate MCP record contains a real local load and no evaluator scores. Review your own saved AgentCore Evaluate export offline with <code>evalarc trace-import</code>.</p></section>
     <div class="facts" aria-label="Audit scope">
       <div><strong>__TASK_PACK_COUNT__</strong><span>task packs</span></div>
       <div><strong>__FAULT_COUNT__</strong><span>declared faults caught</span></div>

site/labs/research.html

@@ -2,7 +2,7 @@
 <meta name="viewport" content="width=device-width,initial-scale=1">
 <title>Research records &#183; EvalArc</title><style>
 :root{color-scheme:dark}*{box-sizing:border-box}body{margin:auto;max-width:1050px;padding:24px;background:#10191c;color:#edf5ec;font:16px/1.6 system-ui}a{color:#b2f4ca}h1{font-size:clamp(2.4rem,6vw,4rem);line-height:1.1}h2{margin-top:45px}table{border-collapse:collapse;width:100%;margin:20px 0}td,th{padding:10px;text-align:left;border-bottom:1px solid #48625e;overflow-wrap:anywhere}section{background:#172428;padding:20px;border-radius:12px;margin:20px 0}code{overflow-wrap:anywhere}.scope{border-left:3px solid #ffb7a8;padding-left:18px}:focus-visible{outline:3px solid #ffb7a8} @media(max-width:480px){body{padding:15px}section{padding:12px}td,th{padding:6px;font-size:14px}}</style>
-<a href="../">&#8592; EvalArc</a> &#183; <a href="../skill-impact/">27-trial Skill Impact Lab</a>
+<a href="../index.html">&#8592; EvalArc</a> &#183; <a href="../skill-impact/index.html">27-trial Skill Impact Lab</a>
 <h1>Keep the failed attempts.<br>Check the artifact.</h1>
 <p>Recorded public development experiments, 14 September 2026. Qwen3-8B/Qwen3-4B on an NVIDIA L40S. No private customer data or personal agent history.</p>
 <p class="scope">These pilots test specific integration paths and inspectable outcomes. They do not establish skill or memory efficacy, a general leakage detector, or native commercial-agent interoperability.</p>

site/labs/skill-impact.html

@@ -12,7 +12,7 @@
 </head>
 <body>
 <a href="#explorer" class="skip">Skip to trial explorer</a>
-<header><a href="../">&#8592; Playground</a><span>Robot Reel &#215; Skills Anywhere &#215; EvalArc</span><a href="METHODS.md">Methods &amp; limits</a></header>
+<header><a href="../index.html">&#8592; Playground</a><span>Robot Reel &#215; Skills Anywhere &#215; EvalArc</span><a href="METHODS.md">Methods &amp; limits</a></header>
 <main>
 <p class="eyebrow">Real model calls / NVIDIA L40S / Public development evidence</p>
 <h1>A skill loaded.<br><em>Did the task pass?</em></h1>

src/evalarc/__init__.py

@@ -1,3 +1,3 @@
 """Auditable evaluations for AI agents."""
 
-__version__ = "0.11.0"
+__version__ = "0.11.1"

tests/test_site.py

@@ -45,6 +45,24 @@ def test_bundle_rejects_changed_evidence_and_extra_files(tmp_path):
     assert "__AUDIT_COVERAGE__" not in (folder / "index.html").read_text()
     assert (folder / "index.html").read_text().count("single-case dependencies") == 3
     assert (folder / "robot/index.html").is_file()
+    from scripts.verify_research import verify_lab, verify_records
+
+    for name, verifier in (("skill-impact", verify_lab), ("research", verify_records)):
+        verifier(folder / name)
+        original = builder.ROOT / "examples" / name
+        for source in original.rglob("*"):
+            if source.is_file() and source.relative_to(original).as_posix() not in (
+                "index.html",
+                "manifest.json",
+            ):
+                assert (folder / name / source.relative_to(original)).read_bytes() == (
+                    source.read_bytes()
+                )
+        page = (folder / name / "index.html").read_text()
+        assert 'href="../index.html"' in page
+        assert 'href="../"' not in page
+    for name in ("skill-impact", "research", "trace-workbench", "trace-mcp"):
+        assert f'href="{name}/index.html"' in (folder / "index.html").read_text()
     assert "single-case dependency" in (folder / "robot/index.html").read_text()
     audit = folder / "support" / "audit.json"
     original = audit.read_bytes()