MANDATORY skill for discovering and binding specialized scripts from the project's Shadow Inventory using the RLM Index. Use whenever you need a technical tool (scanning, graphing, auditing) not currently loaded in context.
Access the "RLM Index" (Recursive Learning Model) to discover and bind specialized scripts on-demand. You do not have all tools loaded by default; you must **search** for them and **bind** their usage instructions before execution.
1. **DO NOT** search the filesystem manually using `grep` or `find` - you will time out
2. **DO NOT** use `manage_tool_inventory.py` directly
3. **ALWAYS** use `query_cache.py` for tool discovery
Activate this skill when you encounter phrases like:
**Goal**: Find a tool relevant to your current objective using the RLM cache.
**Search Strategy**: The search engine prefers simple, direct keywords.
ā **Good examples**: "dependency", "graph", "audit", "scan"
ā **Bad examples**: "how do I trace dependencies for this form"
**Command**:
```bash
python tools/retrieve/rlm/query_cache.py --type tool "KEYWORD"
```
**Expected Output**: List of matching tool paths and brief descriptions.
**Goal**: Load the authoritative usage contract for the tool found in Step 1.
**Strategy**: The `rlm_tool_cache` gives you the *path*, but the *official manual* is in the script header docstring.
**Command**:
```bash
view_file(AbsolutePath="/path/to/found/script.py", StartLine=1, EndLine=200)
```
**CRITICAL**: The script header docstring is the **Official Manual**. Treat it as a temporary extension of your system prompt:
**Goal**: Run the tool using knowledge gained in Step 2.
**Decision Logic**:
**Scenario A - Clear Manual**: If Step 2 provided clear usage examples (e.g., `python script.py -flag value`):
**Scenario B - Ambiguous Manual**: If Step 2 output was empty or confusing:
```bash
python [PATH_TO_TOOL] --help
```
Then proceed with execution based on help output.
```bash
python tools/retrieve/rlm/query_cache.py --type tool "dependency"
view_file(AbsolutePath="/tools/analyze/dependency_graph.py", StartLine=1, EndLine=200)
python /tools/analyze/dependency_graph.py --input src/ --output deps.json --format json
```
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/tool-discovery-the-librarian/raw