{skill_name}/
├── skills/
│ └── {skill_name}/
│ ├── SKILL.md
│ └── references/ (if any)
└── modes/
├── {mode-slug-1}.yaml
├── {mode-slug-2}.yaml
└── ...
Copy skills/ content to ~/.kilocode/skills/ and merge modes/ into .kilocodemodes or ~/.kilocode/custom_modes.yaml.
Automatic mode (no questions):
/claude-workflow-skill-to-kilocode -a workflow-apex
# Creates: ./workflow-apex/skills/... + ./workflow-apex/modes/...
With conversion logs:
/claude-workflow-skill-to-kilocode -a -s utils-refactor
# Creates output + logs in .claude/output/skill-converter/
Convert by path
/claude-workflow-skill-to-kilocode /path/to/my-skill
Auto mode - no questions
/claude-workflow-skill-to-kilocode -a workflow-apex
Save conversion logs
/claude-workflow-skill-to-kilocode -a -s meta-hooks-creator
</examples>
</parameters>
<state_variables>
**Persist throughout all steps:**
| Variable | Type | Description |
|----------|------|-------------|
| `{skill_path}` | string | Full path to Claude Code skill |
| `{skill_name}` | string | Skill name (directory name) |
| `{auto_mode}` | boolean | Skip confirmations |
| `{save_mode}` | boolean | Save conversion logs |
| `{logs_dir}` | string | Conversion logs directory (if save_mode) |
| `{output_root}` | string | Output directory: `./{skill_name}/` |
| `{skill_structure}` | object | Parsed skill frontmatter and body |
| `{referenced_agents}` | list | Subagents found in skill |
| `{converted_modes}` | list | Modes created from agents |
| `{kilocode_skill}` | object | Final converted skill |
</state_variables>
<workflow>
**Conversion flow:**
step-00-init: Parse flags, locate skill ↓ step-01-analyze: Read skill, identify structure, find agent references ↓ step-02-convert-agents: Convert referenced agents to modes (if any) ↓ step-03-convert-skill: Convert skill frontmatter and body ↓ step-04-validate: Validate conversion completeness ↓ step-05-save: Save to Kilocode location
</workflow>
<entry_point>
**FIRST ACTION:** Load `steps/step-00-init.md`
</entry_point>
<step_files>
| Step | File | Purpose |
|------|------|---------|
| 00 | `steps/step-00-init.md` | Parse flags, locate skill, setup state |
| 01 | `steps/step-01-analyze.md` | Analyze skill structure, find agent refs |
| 02 | `steps/step-02-convert-agents.md` | Convert agents to Kilocode modes |
| 03 | `steps/step-03-convert-skill.md` | Convert skill to Kilocode format |
| 04 | `steps/step-04-validate.md` | Validate conversion completeness |
| 05 | `steps/step-05-save.md` | Save and report results |
</step_files>
<execution_rules>
- **Load one step at a time** - progressive loading
- **Persist state variables** across steps
- **Follow next_step directive** at end of each step
- **Save outputs** if `{save_mode}` = true
- **Skip AskUserQuestion** if `{auto_mode}` = true
</execution_rules>
<references>
- [references/frontmatter-mapping.md](references/frontmatter-mapping.md) - Field conversion reference
- [references/tool-to-group-mapping.md](references/tool-to-group-mapping.md) - Claude tools to Kilocode groups
- [references/conversion-patterns.md](references/conversion-patterns.md) - Common conversion patterns
</references>
<success_criteria>
- Skill frontmatter correctly converted
- Body content adapted for Kilocode
- All referenced agents converted to modes
- Dynamic context documented as limitations
- Converted skill validates against Kilocode schema
</success_criteria>