Do I need hooks, agent instructions, or both?
Do I need hooks, agent instructions, or both?
They do different jobs. Hooks report threads and runs and enforce stops, without relying on the model. Instructions tell the model when to search and write Slaunt Context, which hooks can’t do. On Claude Code and Codex use both; elsewhere only instructions are available.
Does Slaunt store my prompts?
Does Slaunt store my prompts?
Only if Chat capture is set up (capture hooks or
chat_capture), and then encrypted. Lifecycle hooks send metadata only, and Context events are one-line summaries. See Security and data.Can one agent use two workspaces?
Can one agent use two workspaces?
No. A token belongs to one workspace; sign the client in again with the other account to switch.
Why does my agent say the work is already done?
Why does my agent say the work is already done?
It found a matching Context event, which is the point: it avoids redoing finished work. If the event is wrong, log a correcting event.