Skills and playbooks

The reusable-procedure catalog. Where an organization's know-how is stored, imported from GitHub, and surfaced by relevance each turn.

Skills and playbooks are the procedures an organization reuses. They share one immutable substrate and two catalog surfaces: /skills and /playbooks.

Two views of one substrate

The distinction is content, not mechanism: a playbook is a skill that reads as a full procedure. Both are versioned and both are activated the same way.

Imported from GitHub

GitHub is the source of an organization’s procedures. From /skills you can import SKILL.md files from a repository with one click. A periodic resync can also sweep the organization’s repositories and re-import new or changed skills through the same path; it is off by default, enabled by an operator-set interval.

• Discovery runs over a server-side shallow clone, resolving the head with git ls-remote.
• Imports are pinned to a commit and idempotent by source, so re-importing never duplicates. The resync skips any repository whose head is unchanged since its last sync on a single git ls-remote.

See Knowledge and learning for the mechanism.

Surfaced by relevance

The catalog the model sees each turn is not a fixed top-N by usage. On every unpinned turn it is re-ranked by relevance to the prompt, so the procedures that match the current ask surface into context. You can also pin a specific skill or playbook from the composer when you want it guaranteed in the turn.