uygnoey Long context capability brief
Instructions
You are assisting a developer or team deciding how to use very long context windows (up to 1M tokens) for Claude Sonnet 4.
Do the following:
- Identify the workload type
- Large-scale code analysis (entire codebases)
- Document synthesis (large sets of contracts/papers/specs)
- Context-aware agents (multi-step workflows across many tool calls)
- Recommend what to include in the prompt
- For codebases: source, tests, and key docs needed to understand architecture.
- For document sets: the full set (or curated subsets) plus a clear synthesis question.
- For agents: relevant tool definitions, key policies/specs, and interaction history.
- Call out operational considerations
- Note that pricing changes apply for prompts over 200K tokens.
- Suggest prompt caching and batch processing when appropriate to reduce latency/cost.
Examples
Example 1: Codebase analysis request
Create a one-page plan describing:
- What files you would include (source/tests/docs)
- What questions you would ask to understand architecture and cross-file dependencies
Example 2: Document synthesis request
Draft a synthesis prompt that:
- States the synthesis goal
- Specifies how to cite evidence from documents
- Requests a coherent output structure
Source
Derived from: https://claude.com/blog/1m-context