The Hub

A skill is a reusable capability Lectern applies for you — either a recorded set of steps or a short learned procedure. The Hub is where you browse community skills and manage your own. Matching skills auto-apply on a task; you can also attach one explicitly with /skill.

Browse & install

The Hub lists skills in tiers — official, curated collections, and community — each with what it does and how many people use it. Installing one adds it to your local skill set so it can match future tasks. You can browse the same catalog on this site.

Create a skill

• Record it. Start a recording, do the task once, and Lectern captures the concrete steps (commands run, files touched). Exploratory commands like ls or git status are filtered out so the recording stays clean.
• Review before saving. Before it's stored you can edit the captured steps — rename it, drop a step, fix a command — so a skill is exactly what you meant, not a raw transcript.
• Or write one. A procedural skill is just its rules and steps in a small file; create it from scratch when there's nothing to record.

A recorded skill replays its steps when applied (deterministic, no re-planning tokens); a procedural skill guides the model instead. Skills are usage-scored, and one that repeatedly fails pauses itself and is surfaced for you to re-enable rather than failing silently.

Import a skill

Have a skill file from someone else? Import it from disk and it joins your set immediately — same as installing from the Hub, but from a local file. Two formats work: Lectern's own .json bundle, and the SKILL.md open standard (.md with frontmatter and a markdown body) that the wider ecosystem publishes — so a skill written for any SKILL.md-compatible agent imports straight into Lectern. Skills are plain, inspectable files; read one before you run it.

Share your own

Export a skill to a shareable bundle, or publish it so others can install it from the Hub. Because a skill can run commands and edit files, Lectern audits a bundle for prompt-injection and unexpected behavior before it's shared — the same scrutiny you should give anything you import.

Safety

• Skills are local files you can read, edit, and delete.
• Imported and published skills are audited for injection and surprising behavior.
• A failing skill self-pauses; nothing keeps re-running a broken procedure.