Skills

The agent's toolset: carry a marketplace skill, drop one, read the trigger index.

1. Step 1 of 3

   Give the agent a skill

   Skills are the agent's toolset, and they add/rm/ls like every other agent collection (no more pin/unpin jargon). add carries a marketplace skill by handle; --why records why THIS agent holds it, in its own words. Add @version to lock a release; leave it off to track latest. The toolset stays curated and bounded on purpose: an agent is composed of the skills you give it, not the whole marketplace.

   plm agent skill add

   $ plm agent skill add edvard/web-verify --why "screenshot before claiming UI done"

   ✓ added edvard/web-verify

   full form · plm agent skill add <owner/slug[@version]> [--why …]

2. Step 2 of 3

   Drop one

   Removes the skill from the agent's foundation. The skill itself is untouched in the marketplace; only this agent stops carrying it. Match by handle — the version suffix is ignored, so you drop it however it was added.

   plm agent skill rm

   $ plm agent skill rm edvard/web-verify

   ✓ removed edvard/web-verify

   full form · plm agent skill rm <owner/slug>

3. Step 3 of 3

   Read the trigger index

   This is the overview the agent reads to know WHICH skill to fire WHEN. Each row shows the skill, its fire-when (derived live from that skill's own when-to-use, so it never drifts), and why the agent carries it. A query filters across ref, fire-when and reason, so `skill ls next` finds every skill about Next.js. The same index is compiled into BOOT.md, so a booted agent already holds the menu without loading a single skill up front.

   plm agent skill ls

   $ plm agent skill ls verify

   edvard/web-verify@1.0.0

   fire when: Use before declaring ANY web page, UI component, or visual change done…

   yours: screenshot and LOOK before claiming UI done

   full form · plm agent skill ls [query]