OpenClaw Skills - Converted from Claude Code

This repository contains a collection of skills and agent specifications converted from Claude Code format to OpenClaw format.

Contents

• skills_tree/: ✓ USE THIS - OpenClaw-ready structure with each skill in its own directory
• skills_flat/: Reference only - All skill files in a flat directory (for browsing)
• skill_manifest.tsv: Mapping between flat files and tree structure
• skill-routing-intent-mapping.md: Central routing map for skill triggers and intent detection

For installation: Use skills_tree/ - it has the proper directory structure OpenClaw expects (skill-name/SKILL.md)

Important: Path Deidentification

All paths and hostnames in these files have been deidentified for public consumption. Before using these skills, you will need to update paths throughout the files to match your environment.

Generic Paths Used

The following generic placeholders are used and need to be replaced:

Generic Path	Description	Replace With
/path/to/source/	Source/mount directory	Your actual source directory
/path/to/venv/	Python virtual environment	Your Python venv path
/path/to/scripts/	User scripts directory	Your scripts directory
/path/to/reports/	Output/reports directory	Your reports directory
/path/to/documents/	Documents directory	Your documents path
/path/to/work/	Work directory	Your work directory
~/.claude	Claude config directory	Usually kept as-is
primary	Primary/master hostname	Your master machine name
host1, host2	Satellite hostnames	Your satellite machine names
hostname	Generic hostname placeholder	Your actual hostname

How to Update Paths

1. Find all path references in your chosen skills:

   grep -r "/path/to/" skills_flat/

2. Replace with your actual paths:

   # Example: Update Python venv path
   find skills_flat/ -type f -name "*.md" -exec sed -i 's|/path/to/venv|/home/yourusername/venv|g' {} +

3. Update hostnames if using config-sync or similar:

   # Replace 'primary' with your master machine name
   sed -i 's/primary/yourmachinename/g' skills_flat/config-sync.SKILL.md

Shebangs

Python scripts reference a generic shebang:

#!/path/to/venv/bin/python

Update this to your actual Python interpreter path:

#!/home/yourusername/venv/bin/python
# or
#!/usr/bin/env python3

Usage with OpenClaw

These skills are designed to be imported into OpenClaw. Each skill file contains:

• YAML frontmatter: Required name and description fields for OpenClaw activation
• Trigger keywords: Keywords that activate the skill
• Instructions: Detailed workflow and behavior specifications
• Context: Domain knowledge and operational guidelines
• Examples: Usage patterns and expected behaviors

⚠️ Critical: AGENTS.md Configuration Required

OpenClaw will NOT route skills effectively without a proper AGENTS.md file in your workspace.

The included skill-routing-intent-mapping.md file provides a basic skill index, but OpenClaw requires an AGENTS.md file in your workspace (~/.openclaw/workspace/AGENTS.md) with explicit routing rules and priority ordering.

Why AGENTS.md is Required

1. Activation: All skills need proper YAML frontmatter (✓ already included in these files)
2. Routing: Default routing is too generic - you need explicit intent mapping and priority order

Without proper routing configuration, OpenClaw may:

• Use generic bundled skills instead of your specialists
• Route requests to the wrong skill
• Miss relevant skills entirely

Setting Up AGENTS.md

Create or update ~/.openclaw/workspace/AGENTS.md with a skill routing section. Here's the recommended structure:

## Skill Routing (Required)

Routing priorities (strict order):

1. Explicit user skill mention wins (`$skill-name` or `use <skill-name>`).
2. If request is multi-step/PRD/end-to-end, invoke `agent-orchestrator` first.
3. Use intent mapping below to choose specialist skills.
4. Only use generic bundled skills when no mapped specialist fits.

Intent mapping (primary picks):

- Implementation/build/code/refactor: `agent-dev-coder` (+ `tdd-workflow` when tests requested).
- Bug/root-cause/debugging: `agent-issue-investigator` + `systematic-debugging`.
- Tests/verification: `agent-validation-agent` + `tdd-workflow`.
- Code review: `agent-code-reviewer`.
- Security/OWASP: `agent-security-specialist` + `security-best-practices`.
- Performance/profiling: `agent-performance-optimizer`.
- Documentation: `agent-documentation-scribe` + `documentation-standards`.
- Python analytics: `agent-python-analytics-specialist` + `python-analytics`.
- Medical imaging/PACS/DICOM: `IT-medical-imaging-informatics`.

[Add mappings for all skills you plan to use]

Disambiguation rules:

- Prefer at most 4 active skills per subtask unless orchestrator expands scope.
- If request includes both implementation and testing, pick both coder + validation.

Forced overrides (must apply):

- If prompt includes `implement`, `build`, `write function`: include `agent-dev-coder`.
- If prompt includes `bug`, `debug`, `regression`: include `agent-issue-investigator`.
- If prompt includes `CI/CD`, `deployment`, `infra`: include `agent-devops-architect`.

Using the Provided Mapping File

The skill-routing-intent-mapping.md file contains descriptions and trigger keywords for all 80 skills. Use it as a reference when building your AGENTS.md routing rules.

Recommended approach:

1. Read through skill-routing-intent-mapping.md to understand available skills
2. Identify the 10-20 skills most relevant to your workflow
3. Add explicit routing rules to your AGENTS.md for those skills
4. Test and refine based on actual usage

Directory Structure Explained

This repository contains two versions of the skills:

• skills_tree/ - OpenClaw-ready structure with each skill in its own directory

  • ✓ Use this for installation
  • Each skill has its own subdirectory with SKILL.md inside
  • Ready to copy directly to ~/.openclaw/skills/

• skills_flat/ - Flat reference structure with all skills in one directory

  • For browsing and reference only
  • NOT compatible with OpenClaw's expected directory structure

Example Workspace Setup

# CORRECT: Copy skills_tree (has proper directory structure)
cp -r skills_tree/* ~/.openclaw/skills/

# OR: Copy individual skills you want
cp -r skills_tree/python-analytics ~/.openclaw/skills/
cp -r skills_tree/tdd-workflow ~/.openclaw/skills/
cp -r skills_tree/agent-dev-coder ~/.openclaw/skills/

# Reference the routing map when building your AGENTS.md
cat skill-routing-intent-mapping.md

# Edit your workspace AGENTS.md
nano ~/.openclaw/workspace/AGENTS.md

# Refresh OpenClaw to discover new skills
# In OpenClaw chat: "refresh skills"
# Or restart: openclaw gateway restart

Important: Do NOT copy from skills_flat/ - OpenClaw requires each skill to be in its own subdirectory with a SKILL.md file inside.

Notes

• Some skills may reference other skills or agents - ensure dependencies are available
• Skills in skills_tree/ maintain the original directory structure for reference
• Skills in skills_flat/ are all in one directory for easier browsing
• The manifest file maps between these two structures

Contributing

If you find issues with the conversion or have improvements, please open an issue or pull request.

License

Please respect the original authorship and licensing of these skills. This is a conversion/export for interoperability purposes.