Skip to content
Knowledge Trust Ladder

Releases

All versionsAll versions forward

Version 0.9.0

Initial release: four Agent Skills that turn a repository’s scattered knowledge into a maintained, trusted LOKF knowledge bundle - built once, kept current, and reviewed by a person, rather than rediscovered every session.

  • lokf-sidecar - bootstraps a fresh .lokf/ sidecar into a repository that has none: tooling, docs, and a dummy skeleton from bundled templates.
  • lokf-librarian - scrapes the repository, derives concepts with their sources, wires typed relationships, audits the bundle against the LOKF schema, and hands off for review. Runs often, including on a schedule; deals in facts about the repository, never in verdicts about truth.
  • lokf-curator - a human curator’s assistant: a one-screen trust and freshness report, and an opt-in review session that records a person’s confirm/correct/retire/send-back verdict directly in the bundle’s frontmatter.
  • lokf-docent, the reader’s entry point. It answers from the bundle first with each concept’s trust label, verifies exact values at the source, and records a gap in .lokf/feedback.md for the librarian when the bundle has no answer. docs/examples/docent.md has real transcripts.

This repository dogfoods its own skills: .lokf/ here is a real bundle built by lokf-sidecar and lokf-librarian, self-describing all four skills, this repository’s own governance, and its CI.