florianbruniaux/claude-cowork-guide

Purpose and Delivery Model

This is a documentation/content repository for Claude Cowork, Anthropic’s desktop assistant, aimed at knowledge workers and small businesses. It distributes guided learning, copy-paste prompts, and operational workflow playbooks rather than executable application code. The primary entry point is README.md, with French parity via README.fr.md.

User Onboarding and Core Guides

The canonical learning sequence introduces Cowork, setup, capabilities, security, and troubleshooting. These documents establish operating boundaries: local folder access, human approval of plans, backup before deletion, and avoiding credentials in accessible folders.

• guide/00-overview.md

• guide/01-getting-started.md

• guide/03-security.md

Workflow Playbook Library

The main actionable content is a catalog of business workflows. Each workflow translates a business objective into a repeatable Cowork interaction, typically involving user-provided local files and generated outputs. The workflow index is the discovery boundary, while individual documents provide domain-specific instructions.

• workflows/README.md

• workflows/file-organization.md

• workflows/invoice-generation.md

English variants are maintained beside the primary workflow files (for example, workflows/file-organization.en.md), indicating a bilingual publication pipeline.

Prompt and Template Assets

Prompt collections provide reusable task instructions for file operations, document creation, extraction/OCR, and research. Persona-oriented Claude configuration examples supplement these with role-specific starting points. These are static consumable assets, not runtime prompt execution code.

• prompts/file-ops.md

• prompts/data-extraction.md

• examples/claude-md-templates/consultant.md

Reference and Interactive Navigation

Reference material supports quick lookup and product evaluation; an onboarding prompt supplies a personalized navigation mechanism within Cowork. This layer connects users from general onboarding to appropriate workflow or prompt assets.

• reference/cheatsheet.md

• reference/faq.md

• tools/onboarding-prompt.md

Machine-Readable Release and Knowledge Surface

Structured YAML and LLM-oriented text expose repository metadata and release information to automation or AI consumers. Root metadata anchors release/version identity.

• machine-readable/reference.yaml

• machine-readable/claude-cowork-releases.yaml

• MANIFEST.yaml

Maintenance, Validation, and Distribution

Shell scripts maintain translation consistency, synchronize version references, and update repository statistics. GitHub Actions invokes translation validation on repository changes. Release-maintenance instructions are available as Claude/Cowork command and agent skill assets.

• scripts/check-translations.sh

• .github/workflows/translation-check.yml

• .claude/commands/cowork/update-releases.md

Runtime and External Boundaries

There is no deployed application runtime. Content is consumed in Claude Desktop on macOS and Windows, requiring Pro or Max subscriptions. Cowork is the execution environment: it receives prompts/files and sends them to Anthropic services; users grant filesystem access and review proposed operations. Documented external integrations include Asana, Canva, Cloudflare, Figma, GitHub, Google Drive, Jira, Linear, Notion, Sentry, and Slack.