Monroe

Documentation

Skills

Shipping in stages. Implicit + explicit skill capture is live today — Monroe remembers corrections in-thread and persists the rule to its memory store. The dedicated Dashboard → Skills tab for inspect/edit/disable/delete + the imported-skills marketplace are on the public roadmap for the next release. Today the way to inspect or clear a skill is to ask Monroe directly in a DM ("what skills do you have for #customer-support?" / "forget that we always include ARR in summaries").

A skill is a learned procedure Monroe remembers across sessions. Think of it as "how Monroe does X for your team specifically."

Examples:

  • "How we write release notes" — after the third time Monroe drafts release notes for your team, it learns the structure you prefer: sections in this order, this tone, link format like that.
  • "Which accounts are tier-1" — Monroe learns the named accounts that always get the deluxe research packet.
  • "How we triage customer feedback" — buckets, priority labels, which channel each one routes to.

How skills get created

Three ways:

1. Implicit — Monroe notices "the user corrected me the same way three times" and writes a skill. Next time, it does it right by default.

2. Explicit — say "remember this: when summarizing #customer-support, always include the customer name and ARR if known." Monroe writes a skill and confirms.

3. Imported — workspace admins can clone skills from a marketplace or another workspace (Enterprise feature, Q3 2026).

Inspecting and editing skills

Dashboard → Skills tab. Every skill has:

  • A name (Monroe assigned, you can rename)
  • A description (what it does)
  • A scope (which channels/tools/contexts it applies to)
  • A history (when it was created, when it was last used)

You can edit the description, narrow the scope, disable a skill, or delete it. Deleting is a one-way operation — Monroe forgets that procedure entirely.

Skills vs system prompt

Skills are different from a system prompt:

  • System prompt is the same for everyone using Monroe.
  • Skills are workspace-specific procedural learning. Your Monroe's skills aren't shared with anyone else's Monroe.

If two skills conflict, the more specific one wins (e.g. a #revenue-only skill beats a workspace-wide skill in #revenue).

Next: run receipt format.