Skip to main content
The server ships 2 code-defined prompts (v1 cap: 2). A prompt only orients the agent — it never queries or persists; arguments arrive as strings over the MCP wire.

​
briefing

Orient yourself at the start of a session: call the briefing TOOL with an explicit selector (no unfiltered default) to surface open/overdue commitments, blockers, stale candidates, recent decisions, and preferences. Title: Session briefing

​
Arguments

{
  "type": "object",
  "properties": {
    "selector": {
      "type": "string",
      "enum": [
        "all",
        "scope",
        "project"
      ],
      "description": "Which slice to brief over: a single scope, a single project, or all memories."
    },
    "mode": {
      "type": "string",
      "enum": [
        "brief",
        "full"
      ],
      "description": "brief (counts + top items, the default) or full (the bounded lists)."
    }
  },
  "required": [
    "selector"
  ]
}

​
debrief

Close a session cleanly: review what happened, then persist what is worth keeping — decisions, commitments, and follow-ups — by calling the remember TOOL for each, and resolve any completed commitments. Title: Session debrief

​
Arguments

{
  "type": "object",
  "properties": {
    "scope": {
      "description": "Optional scope (kebab-case, e.g. work, personal) the session belonged to.",
      "type": "string",
      "minLength": 1,
      "maxLength": 64,
      "pattern": "^[a-z0-9][a-z0-9-]*$"
    }
  }
}