When disclosure_level=0, system prompt only injects skill name + description (summary mode). SkillDetailTool is injected into the tool set, allowing the LLM to load full instructions on-demand via skill_detail(query). This reduces context window consumption when many skills are registered. |
||
|---|---|---|
| .. | ||
| documents | ||
| e2e | ||
| integration | ||
| routes | ||
| tools | ||
| unit | ||
| __init__.py | ||
| conftest.py | ||
| test_routing_chain.py | ||