Introduction to Claude Cowork

4 min read

Source Snapshot

Garden Card

This note is a Quartz-ready onboarding map for Claude Cowork, focused on delegation, projects, connectors, skills, plugins, browser use, and review discipline.

1. Executive Summary

Claude Cowork shifts the mental model from question-and-answer chat to delegated work. The user describes an outcome, names the inputs, and lets Claude plan and execute the work arc across local files, connectors, browser context, and tools.

The course emphasizes smallest useful folders, connector setup, permission modes, projects, skills, plugins, Claude in Chrome, scheduled tasks, and human review of final artifacts.

  • Main idea: Delegation beats step-by-step prompting for real work.

  • Why now: AI assistants are moving into files, apps, browsers, and recurring workflows.

  • Where it applies: Reports, proposals, meeting follow-ups, data synthesis, and cross-app research.

Decision Signal

Give Claude a complete work package: deliverable, inputs, nuance, scope, and review boundary.

2. Key Technical Terms

Use these terms to describe Claude Cowork’s working model.

  • Cowork: A delegated working session where Claude plans and executes toward a deliverable.

  • Project: Scoped workspace with instructions, context, scheduled tasks, and accumulated memory.

  • Connector: Integration that lets Claude search or use external apps such as email, calendar, Slack, Drive, or CRM.

  • Skill: Reusable playbook for a recurring process.

  • Plugin: Packaged set of skills, connectors, and subagents around a role or process.

3. Core Notes

3.1 Problem

Many users treat Cowork like Chat and feed it one small question at a time. That underuses its ability to plan across files, apps, and tools.

  • Unclear deliverables produce vague outputs.

  • Oversized folders create noise and risk.

  • Missing context leads to avoidable clarifying loops.

3.2 Mechanism

A strong Cowork prompt names the deliverable, inputs, date range or folder, relevant nuance, and review expectations.

  • Use the smallest folder that contains the work.

  • Connect only the apps relevant to the task.

  • Interrupt early if the plan goes wrong.

3.3 Evidence

The course organizes Cowork around modules for setup, first tasks, global instructions, projects, skills, plugins, Chrome, scheduling, and dispatch.

  • Projects carry standing context for a stream of work.

  • Skills teach Claude a repeatable process.

  • Claude in Chrome bridges tools without dedicated connectors.

3.4 Boundary

Cowork still requires human review. The file may be finished, but facts, business judgment, permanent actions, and sensitive outputs need inspection.

  • Default to asking before unfamiliar actions.

  • Review final artifacts like a colleague’s draft.

  • Scheduled tasks only work when the required desktop environment is available.

4. Concept Map

Use wikilinks to connect this note into the broader Quartz graph.

flowchart LR
  A["Outcome"] --> B["Inputs"]
  B --> C["Nuance"]
  C --> D["Cowork Plan"]
  D --> E["Tool Execution"]
  E --> F["Artifact"]
  F --> G["Human Review"]
  G --> H["Skill or Project Memory"]

Diagram labels stay in English for rendering consistency and easier reuse across published pages.

5. My Take

Cowork is most useful when the human manages delegation quality. The better the outcome, inputs, and review boundary are defined, the less the user has to stitch work together manually.

  • What changed my thinking: Cowork works best with whole work packages, not isolated prompts.

  • What I may do next: Turn recurring deliverables into projects and skills.

  • What still needs verification: Current plan requirements, connector availability, and permission behavior.

Reuse Path

Convert this note into a first-task prompt checklist for Cowork onboarding.

References

Graph View

Backlinks