Coworker offers two distinct ways to give your AI persistent context: Skills and Memories. Both influence how Coworker responds, but they serve different purposes, work differently under the hood, and are managed in different places. Understanding the distinction helps you get more consistent, personalized, and powerful results.
Skills
A Skill is a reusable, shareable module of instructions that you attach to one or more Agents. When a Skill is attached to an Agent, its instructions are automatically included every time that Agent runs — ensuring consistent behavior across every execution.
Think of a Skill as a building block. Instead of rewriting the same instructions in every Agent you build, you write them once as a Skill and attach them wherever they're needed. Skills can be shared across your workspace, so your team doesn't have to recreate common instructions from scratch.
Key characteristics of Skills
Opt-in per Agent — A Skill only applies to Agents it is explicitly attached to. It does not affect chat sessions or other Agents unless added.
Reusable — One Skill can be attached to many Agents simultaneously.
Shareable — Skills published to the Skills Library are available to anyone on your workspace to browse, preview, and attach to their own Agents.
Always injected when attached — If a Skill is on an Agent, its instructions run every time that Agent executes, without exception.
When to use Skills
Enforcing consistent behavior — A "Communication Style" Skill ensures every Agent writes in your company's tone and format.
Standardizing outputs across teams — A branding guidelines Skill can be created once and shared so every team member's Agent applies the same standards.
Keeping Agent prompts focused — Rather than one long system prompt, break instructions into focused Skills and attach only what's relevant to each Agent.
Reusing expertise — If you've built a great set of instructions for one Agent (e.g., how to summarize a sales call), save it as a Skill and apply it to other Agents.
Where to find Skills
Skills live in the Skills Library. Navigate to Agents → Skills to browse, search, create, and manage Skills in your workspace. From the Agent Builder, scroll to the Skills section of any Agent's configuration to attach Skills directly.
Memories
Memories are persistent pieces of context that are automatically injected into your prompts — no attachment or configuration required per session. They exist so you don't have to re-explain your preferences, working style, or organizational context every time you start a conversation.
There are four types of Memories, each with a different scope and injection behavior:
Memory types
Type | Who it applies to | Injection behavior |
Personal Memory | You only | Always injected into your prompts |
Agent Memory | Specific to an Agent | Always injected for every run of that Agent |
Network Memory | Everyone on your network | Semantically injected — only used when there's a strong match between the memory and the task |
Connector Memory | Everyone on your network | Injected only when that specific connector's tooling is actively used in a prompt |
When to use Memories
Personal preferences — "My preferred name is Will, not William." "When I ask for a summary, always use bullet points."
Organizational context — "We operate on fiscal years. Q1 is March–May." "'Lighthouse' refers to our enterprise product tier."
Connector-specific context — Storing field mappings, custom object definitions, or schema notes for connectors like Salesforce or HubSpot so Coworker can interpret your CRM data accurately.
Repetitive instructions — "Whenever I ask you to create a Jira ticket for a feature request, use the Feature Request (FR) project."
How to create a Memory
You can add a Personal Memory in two ways:
During any conversation, say "Remember that..." and Coworker will save it for you.
Go to your Profile → Personal Memories to view, add, edit, or delete memories at any time.
Memories are prioritized by recency and relevancy, so the more memories you create and the more you use Coworker, the more tailored your experience becomes.
Skills vs. Memories at a glance
| Skills | Memories |
What it is | A reusable module of instructions | Persistent contextual information |
How it's applied | Manually attached to specific Agents | Automatically injected into prompts |
Where it lives | Skills Library (Agents → Skills) | Profile → Personal Memories, or admin-managed for network/connector types |
Who it affects | Only Agents it's attached to | You (Personal), your whole network (Company/Network/Connector) |
Typical use case | Standardizing Agent behavior; reusing instructions across Agents | Personalizing responses; injecting organizational or connector context |
Can be shared | Yes — published to the workspace library | Network and Connector Memories are shared across your network (admin-managed) |
Quick rule of thumb: Use Skills when you want to reuse a set of instructions across multiple Agents. Use Memories when you want Coworker to always know something about you, your team, or your data — without having to say it every time.