Triggers

Triggers describe event-driven automation (hooks) such as running a command or prompt before/after agent actions. Support is narrower than MCP or skills and is often gated for safety.

Feature availability

Triggers appear as a Library type with:

• Event name
• Optional matcher
• Action type (command or prompt)
• Enablement flag
• Timeout and metadata fields as exposed by the editor

If an agent lacks a stable local trigger schema, the matrix marks the combination unsupported or unverified.

Supported agents and scopes

Refer to Supported agents for the full grid. In broad strokes:

• Stronger coverage includes Claude Code, Cursor, Codex (with gating), Devin, GitHub Copilot CLI, Augment, OpenCode (gated), and others.
• Several agents leave triggers unsupported (Aider, Antigravity, Cline skills-adjacent automation, CodeWhale, Hermes, Kilo Code, Zed, and more).
• Some combinations are unverified and should be treated as unavailable.

Trust and safety

Workspace triggers can execute during agent sessions. Sync may require Workspace trust/review before apply. Plugin-backed or cloud-only trigger surfaces are not treated as ordinary file-backed Library documents.

Enablement and synchronization

1. Create or import a trigger.
2. Keep the item enabled only when you intend it to be active.
3. Assign to supported agents/scopes.
4. Review sync carefully: hooks are higher risk than passive markdown.

Disable or remove triggers you do not understand before bulk-applying a shared Library.