Documentation Index
Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt
Use this file to discover all available pages before exploring further.
Usage
Subcommands
grade (default)
Grade a skill’s recent sessions:grade auto
Auto-grade with custom expectations:grade baseline
Establish a baseline score before evolution:| Flag | Description |
|---|---|
--eval-set PATH | Use a specific eval set for baseline |
grade baseline persists the no-skill comparison into SQLite so the dashboard,
skill report, and selftune status can tell whether a skill has been measured
against a real baseline yet.
For brand-new draft packages, prefer selftune create baseline
so the package replay and publish path stays in one surface. grade baseline
is still the baseline command for existing skills you are evolving in place.
Common options
| Flag | Description |
|---|---|
--skill NAME | Skill to grade (required) |
--skill-path PATH | Path to the skill file (for baseline or explicit targeting) |
--expectations "..." | Natural language expectations for grading |
--agent NAME | Agent name for multi-agent setups |