feat: Add repo-local skill support + structure tests

- Add .agents/skills/cli-anything/ for OpenClaw agents
- Add tests/test_skill_layout.py for dual-entrypoint validation
- Update README.md with repo-local skill documentation
- Update .gitignore to include new directories

Implements issue HKUDS#30

Author: OpenClaw

.agents/skills/cli-anything/SKILL.md

@@ -0,0 +1,85 @@
+# cli-anything Skill
+
+Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.
+
+## Overview
+
+This skill enables AI agents to transform any software application into an agent-controllable CLI. It follows the proven cli-anything methodology that has successfully generated CLIs for GIMP, Blender, Inkscape, Audacity, LibreOffice, OBS Studio, and Kdenlive.
+
+## Usage
+
+```bash
+# Build CLI for local source
+cli-anything /path/to/software
+
+# Build CLI from GitHub
+cli-anything https://github.com/org/repo
+```
+
+## Prerequisites
+
+- Python 3.10+
+- click >= 8.0
+- pytest
+
+Install dependencies:
+```bash
+pip install click pytest pyyaml
+```
+
+## Methodology
+
+This skill follows the 7-phase cli-anything methodology:
+
+### Phase 0: Source Acquisition
+- Clone GitHub repos or validate local paths
+- Verify source code structure
+
+### Phase 1: Codebase Analysis
+- Analyze application architecture
+- Map GUI actions to API calls
+- Identify existing CLI tools
+
+### Phase 2: CLI Architecture Design
+- Design command groups matching app domains
+- Plan state model and output formats
+- Create software-specific SOP documents
+
+### Phase 3: Implementation
+- Create directory structure: `agent-harness/cli_anything/<software>/`
+- Implement core modules with Click
+- Add REPL support with JSON output mode
+
+### Phase 4: Test Planning
+- Design unit and E2E test coverage
+- Plan validation scenarios
+
+### Phase 5: Test Implementation
+- Write comprehensive tests
+- Ensure 100% pass rate
+
+### Phase 6: Documentation & Publishing
+- Document commands and usage
+- Prepare for PyPI publishing
+
+## Key Files
+
+| File | Purpose |
+|------|---------|
+| `HARNESS.md` | Complete methodology specification |
+| `commands/cli-anything.md` | Main build command |
+| `commands/validate.md` | Validation command |
+| `commands/test.md` | Testing command |
+| `commands/refine.md` | Refinement command |
+
+## Output
+
+- Stateful CLI with REPL mode
+- JSON output for agent consumption
+- Undo/redo support
+- Full test coverage
+
+## Related
+
+- Plugin workflow: See `../cli-anything-plugin/` for Claude Code plugin
+- Examples: See `../blender/`, `../gimp/`, `../audacity/` for implemented CLIs

.gitignore

@@ -18,6 +18,12 @@
 # Step 3: Allow cli-anything-plugin entirely
 !/cli-anything-plugin/
 
+# Step 3.5: Allow .agents for repo-local skills
+!/.agents/
+
+# Step 3.6: Allow tests directory
+!/tests/
+
 # Step 4: Allow each software dir (top level only)
 !/gimp/
 !/blender/

README.md

@@ -514,6 +514,19 @@ The playbook distills key insights from successfully building all 8 diverse, pro
 /cli-anything <software-name>
 ```
 
+### For Repo-Local Usage (OpenClaw Agents)
+
+This project supports dual-entrypoint design:
+
+1. **Plugin workflow** (see above): Use `cli-anything-plugin/` as a Claude Code plugin
+2. **Repo-local skill**: Use `.agents/skills/cli-anything/` for OpenClaw agents
+
+```bash
+# The repo-local skill at .agents/skills/cli-anything/
+# reuses HARNESS.md and command specs from cli-anything-plugin/
+# No additional installation needed for OpenClaw agents
+```
+
 ### For Generated CLIs
 
 ```bash

tests/test_skill_layout.py

@@ -0,0 +1,87 @@
+"""
+Test suite for CLI-Anything repo structure validation.
+
+This test ensures the dual-entrypoint design (plugin + repo-local skill)
+is properly documented and aligned with README claims.
+"""
+
+import os
+import re
+from pathlib import Path
+
+REPO_ROOT = Path(__file__).parent.parent
+
+
+def test_readme_mentions_plugin_and_skill():
+    """README should mention both plugin and repo-local skill paths."""
+    readme = REPO_ROOT / "README.md"
+    content = readme.read_text()
+    
+    # Check plugin path is documented
+    assert "cli-anything-plugin" in content, "README should mention cli-anything-plugin"
+    
+    # Check skill path is documented (dual-entrypoint)
+    assert ".agents/skills" in content or "repo-local" in content.lower(), \
+        "README should mention repo-local skill path"
+
+
+def test_skill_directory_exists():
+    """Repo-local skill directory should exist."""
+    skill_dir = REPO_ROOT / ".agents" / "skills" / "cli-anything"
+    assert skill_dir.exists(), f"Skill directory should exist at {skill_dir}"
+    assert (skill_dir / "SKILL.md").exists(), "Skill should have SKILL.md"
+
+
+def test_skill_reuses_harness():
+    """Skill should reference the existing HARNESS.md."""
+    skill_md = REPO_ROOT / ".agents" / "skills" / "cli-anything" / "SKILL.md"
+    content = skill_md.read_text()
+    
+    assert "HARNESS.md" in content, "Skill should reference HARNESS.md"
+    assert "cli-anything-plugin" in content, "Skill should mention plugin for shared specs"
+
+
+def test_plugin_directory_exists():
+    """Plugin directory should exist for dual-entrypoint design."""
+    plugin_dir = REPO_ROOT / "cli-anything-plugin"
+    assert plugin_dir.exists(), "cli-anything-plugin directory should exist"
+    assert (plugin_dir / "HARNESS.md").exists(), "Plugin should have HARNESS.md"
+    assert (plugin_dir / "commands").exists(), "Plugin should have commands directory"
+
+
+def test_example_appsocumented():
+    """README should list example applications."""
+    readme = REPO_ROOT / "README.md"
+    content = readme.read_text()
+    
+    examples = ["GIMP", "Blender", "Inkscape", "Audacity"]
+    for example in examples:
+        assert example in content, f"README should mention {example} as example"
+
+
+if __name__ == "__main__":
+    import sys
+    
+    tests = [
+        test_readme_mentions_plugin_and_skill,
+        test_skill_directory_exists,
+        test_skill_reuses_harness,
+        test_plugin_directory_exists,
+        test_example_appsocumented,
+    ]
+    
+    failed = []
+    for test in tests:
+        try:
+            test()
+            print(f"✓ {test.__name__}")
+        except AssertionError as e:
+            print(f"✗ {test.__name__}: {e}")
+            failed.append(test.__name__)
+    
+    if failed:
+        print(f"\n{len(failed)} test(s) failed")
+        sys.exit(1)
+    else:
+        print(f"\nAll {len(tests)} tests passed!")
+        sys.exit(0)