Skip to main content
These playbooks are written for coding agents and agent harnesses. Human-facing guides explain outcomes and product concepts; this section defines operating contracts, mutation boundaries, expected evidence, and stop conditions.

​
Default operating contract

  1. Inspect the installed SelfTune version, supported harnesses, project root, and current skill locations before changing anything.
  2. Prefer read-only inventory, audit, preview, and plan operations first.
  3. Separate local file changes, hook installation, history import, background services, credentials, remote uploads, and publication into distinct consent decisions.
  4. Show the exact package revision, destination, conflicts, and files affected before a mutation.
  5. Preserve receipts and rollback information.
  6. Verify the resulting state instead of assuming a command succeeded.
Never infer permission to upload, publish, replace a skill, connect an account, install hooks, or delete data from permission to inspect the local library.

​
Choose a playbook

Set up and inventory

Install or reuse SelfTune and return a verified inventory.

Manage the Library

Review duplicates, consolidation, quarantine, and restoration.

Scope Skill Sets

Capture or create a Set, plan project changes, and apply only after approval.

Share and distribute

Verify the sealed revision and stop before each remote write.

Improve and validate

Turn an observed problem into a bounded, evidence-backed proposal.
The installed SelfTune SKILL.md remains the version-matched authority for exact commands and platform-specific behavior. Use selftune --help when these docs and the installed version disagree.