The Runtime Page: Skills, Commands & Hooks

Every project runs its coding agents with a runtime — the set of skills, slash commands, and hooks they have available. The Runtime page (in the project sidebar, between Architect and Settings) is where you see and manage that runtime. It has six tabs: Author, Skills, Commands, Hooks, Stage skills, and All drafts.

Author

This is the default tab and the primary way to add to your runtime. Describe the skill, command, or hook you want in plain language and an agent turns it into a well-formed draft, right in the conversation. Each draft appears as a card you open in a side panel to review — accept it and it goes live (and syncs to your devices and to everyone else in the workspace), or send it back for changes. You never have to hand-write a skill file: you talk about what you want and review what the agent produced.

Skills

Skills are reusable capabilities your agents can load — things like "write a blog post," "run an accessibility review," or a framework-specific helper. The Skills tab is where you browse and manage them.

  • Add a skill — search the skills registry and click Add. Trinity installs it into your project and syncs it to every device and every member, so everyone working on the project gets the same skill.
  • Remove a skill — any skill you added can be removed with the trash button. The installed set is listed for removal without needing to search.

Trinity's own core skills (checking, reviewing, debugging, documenting, testing, and more) are always available to your agents; they aren't listed here to manage. The same skill browser appears during onboarding and when you import an existing project, so you can pick skills up front — but you can always come back to add or remove more later.

Commands

Commands are slash commands your agents can run. This tab manages what's installed: edit a command's body, remove it, and see how each one can be invoked — User only, Model only, or User + model — along with where an imported command came from.

Hooks

Hooks are scripts that run automatically at certain points in an agent's work (for example, a check that runs before a commit). This tab lets you enable or disable each hook, edit its body, set when it runs with a per-harness binding editor (Claude, Codex, OpenCode), and remove it.

Hooks are off until you turn them on. A hook that an agent writes for you arrives disabled — it won't run until you explicitly enable it from this tab, and a disabled hook is shown prominently as inert. That's deliberate: a brand-new hook can't start changing how your agents behave just because it was added.

Stage skills

Trinity runs each story through a pipeline of agent roles — analysis, implementation, review, documentation. The Stage skills tab lets you decide which skills each stage offers its agent, so you can point a specific capability at the exact step where it matters. Any skill you've added or imported shows up here alongside Trinity's catalog, assignable per stage. Leave a stage untouched and it inherits the defaults.

All drafts

This tab collects every AI-authored and imported runtime item waiting for your approval, with a count badge whenever drafts are pending. Your agents don't change your runtime directly — when an agent writes a new skill, command, or hook, or when Trinity finds existing skills inside a project you imported, each one lands here as a draft instead of going live.

Each draft shows up as a card with a full diff of what it would add or change:

  • Accept — the skill, command, or hook goes live and syncs to your devices and to every member. (An accepted hook is added disabled — enable it from the Hooks tab when you're ready.)
  • Reject — the draft is discarded and nothing changes.

A multi-file skill (its main file plus any supporting files) is shown as a single card and accepted or rejected as one unit.

Where drafts come from

  • Authoring — ask an agent on the Author tab to create a skill or command; it turns your request into a well-formed draft.
  • Importing — when you import an existing codebase, any skills already living inside it are lifted out and staged here for review, so you can adopt the ones you want. Your original files are never modified.