Troubleshooting
Installation issues
ontoskills install mcp fails
Check:
- Node.js 18+ is available
- The release artifacts for the current version exist
- Your machine can download GitHub release assets
- No proxy or firewall is blocking
github.comandapi.github.com
Error: “Failed to fetch release metadata”
Network connectivity issue. Check your internet connection and try again. If behind a corporate proxy:
export HTTPS_PROXY=http://proxy.example.com:8080ontoskills install mcpError: “Release does not contain asset”
The platform may not be supported. Check available platforms:
# Supported: darwin-arm64, darwin-x64, linux-arm64, linux-x64uname -m && uname -sontoskills install core fails
The compiler requires Python 3.10+:
python3 --versionIf Python is installed but not found:
export PYTHON=/path/to/python3ontoskills install coreStore and package issues
”Package not found”
- Check the package ID spelling
- Run
ontoskills search <query>to discover available packages - If using a third-party store, verify it’s configured:
ontoskills store listStore skill does not appear in OntoMCP
Skills are enabled by default on install. If a skill is not visible:
- Check if it was disabled:
ontoskills list-installed- Re-enable if needed:
ontoskills enable mareasw/greeting/hello- Rebuild the index:
ontoskills rebuild-index- Restart the MCP process
”Skill still not visible after enable”
The MCP server caches the ontology index. Ensure:
- Index was rebuilt:
ontoskills rebuild-index - MCP server was restarted (close and reopen your AI client)
- Check
~/.ontoskills/ontologies/system/index.enabled.ttlexists
Compiler issues
Compiler commands fail
Install the compiler first:
ontoskills install coreThen initialize the ontology foundation:
ontoskills init-core“ANTHROPIC_API_KEY not set”
The compiler requires an Anthropic API key for LLM-based knowledge extraction:
export ANTHROPIC_API_KEY="your-key-here"ontoskills compile my-skillAdd to your shell profile (~/.bashrc, ~/.zshrc) to persist.
”SHACL validation failed”
Your skill is missing required fields. Check:
- At least one
resolvesIntent(in “When To Use” section) - A clear one-line nature statement at the top
- Proper YAML frontmatter with
nameanddescription
Run with verbose output for details:
ontoskills compile my-skill -v“Nature not extracted”
Add a clear one-line summary at the beginning of your SKILL.md:
# Skill Title
A brief description of what this skill does.
## What It Does...“Missing resolvesIntent”
Ensure your skill has a “When To Use” or similar section:
## When To Use
Use this skill when the user wants to extract text from PDF files.“Skipping embedding generation”
Embedding generation is optional. The compiler skips it with a warning when ontocore[embeddings] is not installed. To enable semantic search vectors:
pip install ontocore[embeddings]“Skill has no declared intents”
Every skill must declare at least one intent for semantic search. Add an intents section to your SKILL.md:
## When To Use
- edit spreadsheet- perform calculationImport issues
Imported source repo compiled, but the skill still is not visible
Imported skills are enabled by default. If not visible:
- Rebuild the index:
ontoskills rebuild-index-
Restart the MCP process
-
If still not visible and the skill was previously disabled, re-enable it:
ontoskills enable <qualified-id>ontoskills rebuild-index“Source import failed”
Ensure:
- Git is installed and accessible
- The repository URL is correct and accessible
- OntoCore is installed:
ontoskills install core
”No SKILL.md files found”
The import process looks for SKILL.md files in the repository. Ensure:
- Files are named exactly
SKILL.md(case-sensitive) - Files are in the repository root or subdirectories
- Files have valid YAML frontmatter
MCP connection issues
”MCP server not starting”
Check the binary exists and is executable:
ls -la ~/.ontoskills/bin/ontomcpIf missing, reinstall:
ontoskills install mcp“Connection refused” or “Timeout”
The MCP server may be slow to start. Check:
- Server is running:
ps aux | grep ontomcp - No port conflicts
- Sufficient system resources
”Claude Code cannot find ontomcp”
Ensure ~/.ontoskills/bin is in your PATH, or use the full path in your MCP configuration:
{ "command": "/home/user/.ontoskills/bin/ontomcp"}“ONNX Runtime shared library not found”
The MCP server uses ONNX Runtime for semantic search. If the library is not in your system path:
# Find the library (installed via pip) — platform-agnosticpython3 -c "import onnxruntime; import os; print(os.path.join(os.path.dirname(onnxruntime.__file__), 'capi'))"Then set the path to the shared library inside that directory:
| Platform | Library name |
|---|---|
| Linux | libonnxruntime.so |
| macOS | libonnxruntime.dylib |
| Windows | onnxruntime.dll |
# Set the path (example for Linux)export ORT_DYLIB_PATH=/path/to/libonnxruntime.soAdd to your shell profile (~/.bashrc, ~/.zshrc) to persist.
Index and state issues
”Index corrupted”
Rebuild from scratch:
ontoskills rebuild-indexIf that fails, check the lock file:
cat ~/.ontoskills/state/registry.lock.json“State files missing”
The state directory should contain:
registry.sources.json— configured storesregistry.lock.json— installed packages
If missing, they’ll be recreated on next operation.
”Permission denied” errors
Check ownership of ~/.ontoskills/:
ls -la ~/.ontoskills/Fix permissions if needed:
chmod -R u+rw ~/.ontoskills/Diagnostic tools
ontoskills doctor
Run a comprehensive health check:
ontoskills doctorThis checks:
- MCP binary exists and is executable
- Core ontology is valid
- Environment variables are set
- Index is consistent
- Available updates
Verbose output
Most commands support -v for detailed logging:
ontoskills compile my-skill -vontoskills install mcp -vReset and recovery
Reset everything
To remove the entire managed home:
ontoskills uninstall --allWarning: This deletes everything under ~/.ontoskills/.
Reinstall from scratch
ontoskills uninstall --allontoskills install mcpontoskills install core # if neededontoskills init-core # if needed# Re-install skillsontoskills install mareasw/greeting/helloGetting help
If your issue isn’t covered here:
- Run
ontoskills doctorand check the output - Search existing issues on GitHub
- Open a new issue with:
ontoskills doctoroutput- Command that failed
- Error message
- Your OS and version