fix: resolve evaluation serialization, benchmark metrics, encoding, and script stability issues

skills/claude-api/SKILL.md

@@ -2,8 +2,8 @@
 name: claude-api
 description: |-
   Reference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration.
-  TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens).
-  SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file).
+  TRIGGER — read BEFORE opening target file: prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens).
+  SKIP only when another provider is being worked on: OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over project hits (run grep FIRST if no provider named).
 license: Complete terms in LICENSE.txt
 ---
 

skills/docx/scripts/accept_changes.py

@@ -7,13 +7,14 @@
 import logging
 import shutil
 import subprocess
+import tempfile
 from pathlib import Path
 
 from office.soffice import get_soffice_env
 
 logger = logging.getLogger(__name__)
 
-LIBREOFFICE_PROFILE = "/tmp/libreoffice_docx_profile"
+LIBREOFFICE_PROFILE = str(Path(tempfile.gettempdir()) / "libreoffice_docx_profile")
 MACRO_DIR = f"{LIBREOFFICE_PROFILE}/user/basic/Standard"
 
 ACCEPT_CHANGES_MACRO = """<?xml version="1.0" encoding="UTF-8"?>
@@ -58,7 +59,7 @@ def accept_changes(
     cmd = [
         "soffice",
         "--headless",
-        f"-env:UserInstallation=file://{LIBREOFFICE_PROFILE}",
+        f"-env:UserInstallation={Path(LIBREOFFICE_PROFILE).as_uri()}",
         "--norestore",
         "vnd.sun.star.script:Standard.Module1.AcceptAllTrackedChanges?language=Basic&location=application",
         str(output_path.absolute()),
@@ -76,7 +77,7 @@ def accept_changes(
     except subprocess.TimeoutExpired:
         return (
             None,
-            f"Successfully accepted all tracked changes: {input_file} -> {output_file}",
+            f"Error: LibreOffice timed out while processing: {input_file}",
         )
 
     if result.returncode != 0:

skills/docx/scripts/office/soffice.py

@@ -17,6 +17,7 @@
 
 import contextlib
 import os
+import platform
 import socket
 import subprocess
 import tempfile
@@ -51,11 +52,13 @@ def run_soffice(args: Iterable[str], **kwargs) -> subprocess.CompletedProcess:
 
 
 def _needs_shim() -> bool:
+    if platform.system() != "Linux":
+        return False
     try:
         s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
         s.close()
         return False
-    except OSError:
+    except (OSError, AttributeError):
         return True
 
 

skills/mcp-builder/scripts/evaluation.py

@@ -83,6 +83,25 @@ def extract_xml_content(text: str, tag: str) -> str | None:
     return matches[-1].strip() if matches else None
 
 
+def _serialize_tool_result(tool_result: Any) -> str:
+    """Serialize MCP tool execution results into clean text for Claude."""
+    if hasattr(tool_result, "content"):
+        tool_result = getattr(tool_result, "content")
+    if isinstance(tool_result, list):
+        parts = []
+        for block in tool_result:
+            if hasattr(block, "text"):
+                parts.append(getattr(block, "text") or "")
+            elif isinstance(block, dict) and "text" in block:
+                parts.append(str(block.get("text") or ""))
+            else:
+                parts.append(str(block))
+        return "\n".join(parts)
+    elif isinstance(tool_result, dict):
+        return json.dumps(tool_result, indent=2)
+    return str(tool_result)
+
+
 async def agent_loop(
     client: Anthropic,
     model: str,
@@ -114,7 +133,7 @@ async def agent_loop(
         tool_start_ts = time.time()
         try:
             tool_result = await connection.call_tool(tool_name, tool_input)
-            tool_response = json.dumps(tool_result) if isinstance(tool_result, (dict, list)) else str(tool_result)
+            tool_response = _serialize_tool_result(tool_result)
         except Exception as e:
             tool_response = f"Error executing tool {tool_name}: {str(e)}\n"
             tool_response += traceback.format_exc()
@@ -144,10 +163,11 @@ async def agent_loop(
         )
         messages.append({"role": "assistant", "content": response.content})
 
-    response_text = next(
-        (block.text for block in response.content if hasattr(block, "text")),
-        None,
-    )
+    text_blocks = [
+        block.text for block in response.content
+        if getattr(block, "type", None) == "text" or hasattr(block, "text")
+    ]
+    response_text = "".join(text_blocks).strip() if text_blocks else None
     return response_text, tool_metrics
 
 

skills/pdf/scripts/convert_pdf_to_images.py

@@ -7,6 +7,7 @@
 
 
 def convert(pdf_path, output_dir, max_dim=1000):
+    os.makedirs(output_dir, exist_ok=True)
     images = convert_from_path(pdf_path, dpi=200)
 
     for i, image in enumerate(images):

skills/pdf/scripts/extract_form_field_info.py

@@ -45,7 +45,7 @@ def make_field_dict(field, field_id):
 
 
 def get_field_info(reader: PdfReader):
-    fields = reader.get_fields()
+    fields = reader.get_fields() or {}
 
     field_info_by_id = {}
     possible_radio_names = set()
@@ -110,7 +110,7 @@ def sort_key(f):
 def write_field_info(pdf_path: str, json_output_path: str):
     reader = PdfReader(pdf_path)
     field_info = get_field_info(reader)
-    with open(json_output_path, "w") as f:
+    with open(json_output_path, "w", encoding="utf-8") as f:
         json.dump(field_info, f, indent=2)
     print(f"Wrote {len(field_info)} fields to {json_output_path}")
 

skills/pptx/scripts/office/soffice.py

@@ -47,15 +47,19 @@ def run_soffice(args: Iterable[str], **kwargs) -> subprocess.CompletedProcess:
 
 
 
+import platform
+
 _SHIM_SO = Path(tempfile.gettempdir()) / "lo_socket_shim.so"
 
 
 def _needs_shim() -> bool:
+    if platform.system() != "Linux":
+        return False
     try:
         s = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
         s.close()
         return False
-    except OSError:
+    except (OSError, AttributeError):
         return True
 
 

skills/skill-creator/eval-viewer/generate_review.py

@@ -276,7 +276,8 @@ def generate_html(
     if benchmark:
         embedded["benchmark"] = benchmark
 
-    data_json = json.dumps(embedded)
+    # Prevent </script in embedded data from prematurely closing inline script tag
+    data_json = json.dumps(embedded).replace("</", "<\\/")
 
     return template.replace("/*__EMBEDDED_DATA__*/", f"const EMBEDDED_DATA = {data_json};")
 

skills/skill-creator/eval-viewer/viewer.html

@@ -1097,9 +1097,13 @@ <h2>Review Complete</h2>
     }
 
     function escapeHtml(text) {
-      const div = document.createElement("div");
-      div.textContent = text;
-      return div.innerHTML;
+      if (text == null) return "";
+      return String(text)
+        .replace(/&/g, "&amp;")
+        .replace(/</g, "&lt;")
+        .replace(/>/g, "&gt;")
+        .replace(/"/g, "&quot;")
+        .replace(/'/g, "&#39;");
     }
 
     // ---- View switching ----
@@ -1129,9 +1133,9 @@ <h2>Review Complete</h2>
       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(String(metadata.runs_per_configuration || "?")) + " runs per configuration";
       html += "</p>";
 
       // Summary table
@@ -1225,7 +1229,7 @@ <h2>Review Complete</h2>
               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>" + escapeHtml(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>";
               if (hasTime) html += "<td>" + (r.time_seconds != null ? r.time_seconds.toFixed(1) : "—") + "</td>";
@@ -1238,7 +1242,7 @@ <h2>Review Complete</h2>
             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) {

skills/skill-creator/scripts/aggregate_benchmark.py

@@ -85,10 +85,13 @@ def load_run_results(benchmark_dir: Path) -> dict:
 
     for eval_idx, eval_dir in enumerate(sorted(search_dir.glob("eval-*"))):
         metadata_path = eval_dir / "eval_metadata.json"
+        eval_name = None
         if metadata_path.exists():
             try:
-                with open(metadata_path) as mf:
-                    eval_id = json.load(mf).get("eval_id", eval_idx)
+                with open(metadata_path, encoding="utf-8") as mf:
+                    meta = json.load(mf)
+                    eval_id = meta.get("eval_id", eval_idx)
+                    eval_name = meta.get("eval_name")
             except (json.JSONDecodeError, OSError):
                 eval_id = eval_idx
         else:
@@ -117,7 +120,7 @@ def load_run_results(benchmark_dir: Path) -> dict:
                     continue
 
                 try:
-                    with open(grading_file) as f:
+                    with open(grading_file, encoding="utf-8") as f:
                         grading = json.load(f)
                 except json.JSONDecodeError as e:
                     print(f"Warning: Invalid JSON in {grading_file}: {e}")
@@ -126,6 +129,7 @@ def load_run_results(benchmark_dir: Path) -> dict:
                 # Extract metrics
                 result = {
                     "eval_id": eval_id,
+                    "eval_name": eval_name or f"Eval {eval_id}",
                     "run_number": run_number,
                     "pass_rate": grading.get("summary", {}).get("pass_rate", 0.0),
                     "passed": grading.get("summary", {}).get("passed", 0),
@@ -139,7 +143,7 @@ def load_run_results(benchmark_dir: Path) -> dict:
                 timing_file = run_dir / "timing.json"
                 if result["time_seconds"] == 0.0 and timing_file.exists():
                     try:
-                        with open(timing_file) as tf:
+                        with open(timing_file, encoding="utf-8") as tf:
                             timing_data = json.load(tf)
                         result["time_seconds"] = timing_data.get("total_duration_seconds", 0.0)
                         result["tokens"] = timing_data.get("total_tokens", 0)
@@ -205,8 +209,13 @@ def aggregate_results(results: dict) -> dict:
 
     # Calculate delta between the first two configs (if two exist)
     if len(configs) >= 2:
-        primary = run_summary.get(configs[0], {})
-        baseline = run_summary.get(configs[1], {})
+        primary_keys = [c for c in configs if c in ("with_skill", "new_skill")]
+        baseline_keys = [c for c in configs if c in ("without_skill", "old_skill")]
+        primary_name = primary_keys[0] if primary_keys else configs[0]
+        remaining = [c for c in configs if c != primary_name]
+        baseline_name = baseline_keys[0] if baseline_keys and baseline_keys[0] != primary_name else (remaining[0] if remaining else configs[1])
+        primary = run_summary.get(primary_name, {})
+        baseline = run_summary.get(baseline_name, {})
     else:
         primary = run_summary.get(configs[0], {}) if configs else {}
         baseline = {}
@@ -237,6 +246,7 @@ def generate_benchmark(benchmark_dir: Path, skill_name: str = "", skill_path: st
         for result in results[config]:
             runs.append({
                 "eval_id": result["eval_id"],
+                "eval_name": result.get("eval_name", f"Eval {result['eval_id']}"),
                 "configuration": config,
                 "run_number": result["run_number"],
                 "result": {
@@ -374,13 +384,13 @@ def main():
     output_md = output_json.with_suffix(".md")
 
     # Write benchmark.json
-    with open(output_json, "w") as f:
+    with open(output_json, "w", encoding="utf-8") as f:
         json.dump(benchmark, f, indent=2)
     print(f"Generated: {output_json}")
 
     # Write benchmark.md
     markdown = generate_markdown(benchmark)
-    with open(output_md, "w") as f:
+    with open(output_md, "w", encoding="utf-8") as f:
         f.write(markdown)
     print(f"Generated: {output_md}")
 

skills/skill-creator/scripts/package_skill.py

@@ -3,18 +3,21 @@
 Skill Packager - Creates a distributable .skill file of a skill folder
 
 Usage:
-    python utils/package_skill.py <path/to/skill-folder> [output-directory]
+    python -m scripts.package_skill <path/to/skill-folder> [output-directory]
 
 Example:
-    python utils/package_skill.py skills/public/my-skill
-    python utils/package_skill.py skills/public/my-skill ./dist
+    python -m scripts.package_skill skills/public/my-skill
+    python -m scripts.package_skill skills/public/my-skill ./dist
 """
 
 import fnmatch
 import sys
 import zipfile
 from pathlib import Path
-from scripts.quick_validate import validate_skill
+try:
+    from scripts.quick_validate import validate_skill
+except ImportError:
+    from quick_validate import validate_skill
 
 # Patterns to exclude when packaging skills.
 EXCLUDE_DIRS = {"__pycache__", "node_modules"}
@@ -110,10 +113,10 @@ def package_skill(skill_path, output_dir=None):
 
 def main():
     if len(sys.argv) < 2:
-        print("Usage: python utils/package_skill.py <path/to/skill-folder> [output-directory]")
+        print("Usage: python -m scripts.package_skill <path/to/skill-folder> [output-directory]")
         print("\nExample:")
-        print("  python utils/package_skill.py skills/public/my-skill")
-        print("  python utils/package_skill.py skills/public/my-skill ./dist")
+        print("  python -m scripts.package_skill skills/public/my-skill")
+        print("  python -m scripts.package_skill skills/public/my-skill ./dist")
         sys.exit(1)
 
     skill_path = sys.argv[1]