Knowledge Agent Skills

Skills are configurable behavior modes on a Knowledge Agent. Each skill defines how the agent responds to a specific type of question or workflow - answering quick questions, conducting deep research, walking through a troubleshooting process, or creating a support ticket.

Instead of building multiple Knowledge Agents for different tasks, you can add multiple skills to a single agent. The agent reads each skill's name and description, then automatically routes to the best-fit skill based on the user's message. Users don't need to manually select a skill - the routing happens behind the scenes.

Skills work everywhere Knowledge Agents work: the Guru web app, Slack, the browser extension, API, and MCP.

---

Default skills

New Knowledge Agents are created with two pre-configured skills:

• Answer. Handles direct questions using your connected sources. AI Agent Center tracking is enabled by default.
• Research. Performs deep, multi-source research for complex questions. AI Agent Center tracking is disabled by default.

Both default skills are fully editable. You can rename them, update their instructions, change their tracking settings, or delete them entirely. They are starting points - not locked templates.

---

How to set up skills

Creating a skill

1. Open your Knowledge Agent and go to the Skills tab.
2. Click Add Skill.
3. Give the skill a name that describes its purpose (e.g., "IT Troubleshooting" or "Benefits FAQ").
4. Write a description that explains when this skill should be used. The agent uses this description to decide whether to route a question to this skill, so be specific about what types of questions or tasks it handles.
5. Add instructions to define how the skill should behave. This is where you control the skill's tone, response format, constraints, and workflow steps. Learn more about writing a strong skill here.
6. Configure the AI Agent Center tracking toggle (see "Tracking skills in the AI Agent Center" below).
7. Click Save.

Creating a skill with Operator Mode

1. Toggle on Operator Mode in your skills-quipped Knowledge Agent chat or from the Skills tab in Agent Settings
2. Click Edit/Add Skill and follow the chat prompts
3. Describe what you want the skill to do and how it should behave. Use the prompts in the Skills Library to make it even easier!
4. Operator Mode will create a draft of the skill for your approval before creating the skill for you

Manual editing in the Skills UI remains available and does not consume credits.

Skills-enabled agents also support automations, which let you schedule any prompt to run on a recurring basis.

Viewing a skill's version history

Every save to a skill creates a new version, and each skill now shows a last edited date. Open the skill's more options menu and select History to open a table of every version — who changed it and when.

From there, you can open any version to view its full content, compare two versions side by side, or revert to a previous version and add a change note explaining why.

📘 Best practices

Unlike Guru Cards, skills and system prompts don't have a draft state — every save publishes immediately as a new version. If you're iterating on a skill's instructions, save smaller, more frequent changes so your version history stays easy to compare.

Access follows the same permission as creating, editing, or deleting the skill — see Access Required above.

---

Tracking skills in the AI Agent Center

Each skill has an independent tracking toggle that controls whether its invocations appear in the AI Agent Center.

• Tracking on: When the skill is invoked, Guru runs post-processing to extract citations, generate an explanation, and classify the interaction as answered or unanswered. A record appears in the AI Agent Center's Questions tab.
• Tracking off: The skill runs normally, but no AI Agent Center question record is created.

Multiple skills on the same Knowledge Agent can each have tracking enabled, and each invocation produces an independent AI Agent Center record.

---

Using built-in tools in skills

Skills can invoke built-in tools to bring additional context into a response. You reference a tool in a skill's instructions — there's no separate toggle or configuration step.

User profile tool

The user profile tool fetches the asking user's employee profile at the time of the conversation. When a skill calls it, the agent receives:

• Name
• Email
• Job title
• Work team
• Work location

Fields are pulled from the user's Guru profile, or from an HRIS sync if your workspace has one configured. The tool returns only the fields that are populated — if a user hasn't filled out their profile, the agent works with what's available.

A typical skill instruction might read: "Get the user's profile to determine where they work, then search for benefits content scoped to their country."

---

Restricting skill access

By default, every skill on a Knowledge Agent is available to all users who can access the agent. KA Owners can optionally restrict a skill so that only specific users or groups can execute it.

1. Open your Knowledge Agent and go to the Skills tab.
2. Select the skill you want to restrict (or create a new one).
3. In the skill's access settings, choose Restrict to specific users or groups.
4. Select the users or groups who should have access. The picker is scoped to users and groups already assigned on the Knowledge Agent's Permissions tab.
5. Click Save.

Restricted skills display a lock indicator in the Skills tab so you can tell at a glance which skills have access constraints applied. Access can also be configured during skill creation — it's surfaced as part of the creation flow, not buried in settings.

At runtime, restricted skills are completely invisible to users who aren't permitted to run them. The agent won't mention, suggest, or reference a restricted skill for unauthorized users — the skill simply doesn't exist from their perspective.

---

Frequently asked questions about Skills

How does the agent decide which skill to use? The agent reads each skill's name and description, then matches the user's message to the best-fit skill using natural-language routing. Users don't manually select skills.

Can I have multiple skills with tracking enabled? Yes. Each tracked skill produces its own independent AI Agent Center record.

Do skills work in Slack? Yes. When a user interacts with a skills-enabled Knowledge Agent in Slack, the agent routes through skills the same way it does in the web app. The @Guru mention is still required for each message in a thread.

Do skills work through MCP and the API? Yes. External tools connecting to Guru through the MCP server or API interact with skills-enabled agents the same way — the agent routes to the appropriate skill automatically.

Can I delete the default Answer or Research skill? Yes. Both default skills are fully editable and deletable. You can also create entirely new skills to replace them.

What happens if I turn tracking off on a skill? Future invocations of that skill won't create AI Agent Center records. Existing records from previous invocations are not affected.

What happens if a user's permissions change while they're in a conversation?

The change takes effect on the next execution. If a user loses access to a restricted skill mid-session, the skill becomes unavailable on their next message — not retroactively within the current conversation.

Does restricting a skill affect credit consumption?

No. Skill permissions are a configuration feature. No credits are consumed when configuring or enforcing access restrictions.