Skills
A skill is a reusable recipe — a step-by-step instruction set that an agent can invoke for a specific class of task. Skills let you teach EM Copilot how you like to do things (“how I prep for a 1:1”, “how I draft a perf review”) so the agent runs the same playbook every time.
Skills are attached to agents (many-to-many — one skill can power several agents) and are matched against the user’s message via trigger phrases. When the agent hears a matching phrase, it runs the skill’s body as a recipe.
How skills work
Section titled “How skills work”- A skill is only available through an agent. Attaching a skill to an agent makes it part of that agent’s toolkit.
- When you send a message to the agent, the agent compares the message against each attached skill’s trigger phrases. A match means the skill’s body becomes the procedure for that turn.
- A skill’s body is plain markdown — typically numbered steps that reference workspace data (members, work logs, feedback, issues) through the EM Kit MCP server.
Built-in skills
Section titled “Built-in skills”Every workspace ships with four built-in skills, pre-attached to the agents that benefit from them:
| Skill | Attached to | What it does |
|---|---|---|
| 1:1 prep | Coach | Pulls recent work logs, open issues, and feedback for a team member into a 1:1 brief. |
| Performance review draft | Coach, Writer | Drafts a performance review from the review period’s data (objectives, feedback, achievements). |
| Team health check | Coach, Researcher | Spots turnover, attrition, and engagement risk signals across the team. |
| Career positioning | Career Advisor | Assesses each team member’s career trajectory against their level. |
Built-in skills can be edited and archived but never deleted.
Configure skills
Section titled “Configure skills”Open Settings → Skills to manage skills.
Create a new skill
Section titled “Create a new skill”Click New skill to open the form:
- Name — Short label shown in the agent form’s skill picker.
- Description — One-line “when this skill should kick in”.
- Triggers — One trigger phrase per line. Whenever an agent that has this skill attached sees a message matching one of these phrases, the skill fires. Use natural language (“prep me for a 1:1”, “talking points for a one-on-one”).
- Body — Markdown editor. The recipe the agent should follow when the skill fires. Numbered steps and explicit tool references work best.
Manage existing skills
Section titled “Manage existing skills”Each row in the table has actions:
- Edit — Open the form. The form shows a Used by panel listing every agent the skill is attached to.
- Archive / Restore — Hide a skill without deleting it. Archived skills do not appear in agent pickers.
- Delete — Permanently remove and detach from any agent that referenced it. Built-in skills cannot be deleted.
Where skills are used
Section titled “Where skills are used”Skills always surface through whichever agent is active. The three places agents are used are therefore the three places skills can fire.
1. EM Copilot chat
Section titled “1. EM Copilot chat”When the agent picked in the draft bar has skills attached, those skills fire automatically as you chat. Saying “prep me for a 1:1 with Sam” to the Coach agent triggers the 1:1 prep skill, which then pulls the relevant data from the workspace and produces the brief.
2. Automations
Section titled “2. Automations”An automation inherits its agent’s full skill set. A daily automation pointed at the Coach agent with a prompt like “Draft 1:1 prep for everyone with a 1:1 today” runs the 1:1 prep skill on every scheduled execution — same recipe, no manual steps.
3. Questions on issues
Section titled “3. Questions on issues”When you mention an agent on an issue with Ask Copilot, the agent’s skills are available while it answers. Mentioning the Career Advisor on an issue about a team member’s growth plan can invoke the Career positioning skill in-thread.
Writing good skills
Section titled “Writing good skills”A few rules of thumb that make skills reliable:
- Be specific in triggers. Add several variations of the phrase you’d actually type. Vague triggers like “help” cause the skill to fire too often.
- Write the body as a procedure. Numbered steps with explicit tool calls beat freeform prose. The agent will follow it more literally than you expect.
- Reference real workspace data. Skills become powerful when they pull from the EM Kit MCP server — point at members, work logs, feedback, or issues by name.
- Keep skills focused. One skill per task type. If a skill does two things, split it — the trigger phrases will be cleaner.