principle-guard-the-context-window

Route verbose outputs, screenshots, and large documents to subagents so the main context receives only summaries. Read files selectively based on relevance, and cap the number of files per phase with explicit turn budgets.

Use subagents whenever a payload is large and only its conclusions matter, such as logs, screenshots, or bulk documents. The subagent processes the raw data and returns a summary, keeping the main context window free for reasoning.

No, templates and references used on every invocation belong inline in the skill file. Storing them separately forces a file read each time, which repeatedly costs context tokens for content that is always needed.

Context overflow degrades reasoning quality, creates compression artifacts, and can halt progress entirely. The context window is finite and non-renewable within a session, so every token spent should justify its cost.

Summarization through subagents loses fine detail, so tasks needing exact raw data must keep it in the main thread. Phase caps and turn budgets also require upfront planning and may not fit highly exploratory work.