Claude Projects provide persistent context and knowledge files for every conversation in a project. These exercises test your understanding of how Projects work, their limitations, and how they differ from API-based context management.
0 / 5 completed
1 / 5
At standup, a colleague asks what Projects in Claude.ai provide that a regular conversation does not. What is correct?
Projects in Claude.ai provide a persistent private context: you define a custom system prompt and upload knowledge files (PDFs, docs, code) that are automatically injected into every new conversation in that project. This eliminates the need to re-paste context each session and keeps the AI's persona and knowledge consistent across all conversations.
2 / 5
During a PR review, a teammate asks how the system prompt in a Project differs from a system prompt in an individual conversation. What is accurate?
A project-level system prompt is configured once and automatically prepended to every conversation created within that project, providing consistent instructions and persona across all sessions. A per-conversation system prompt (set via the API or a one-off chat) only affects that single conversation. This distinction is important for teams building products where consistent AI behaviour is required across many interactions.
3 / 5
In a design review, the team discusses uploading knowledge base files to a Claude Project. Which statement is correct?
Knowledge base files uploaded to a Claude Project are included in the context window of each conversation — they are not fine-tuned into the model or stored in a separate vector database. Claude reads them directly as part of the prompt. This means large files consume context tokens, so the total size of project files must fit within the model's context window alongside the conversation.
4 / 5
An incident report shows that a team's Project stopped accepting new file uploads. A senior engineer asks what the project-level storage limit is. What is the right answer?
Claude Projects have a 200MB storage limit per project for uploaded files. Once reached, new uploads are rejected until existing files are removed. This is a storage-level cap independent of the model's context window limit. Teams managing large knowledge bases need to curate and prune files actively to stay within the limit.
5 / 5
During a code review, a senior engineer asks whether Project context is accessible via the Anthropic API or only through Claude.ai. What is correct?
Projects are a Claude.ai product feature and are not directly accessible via the Anthropic Messages API. API users must manage persistent context themselves — storing system prompts in their application, injecting knowledge as context, and maintaining conversation history. The API does have a Workspaces feature for organisational billing, but that is separate from the Projects product.