Context Compression

When your agent sessions blow past context limits or you're working with multi-million token codebases, this breaks down three compression approaches with real tradeoffs. The key insight is optimizing for tokens per task, not tokens per request, because aggressive compression that loses file paths or error messages costs more in re-fetching than it saves. Anchored iterative summarization with structured sections (intent, files modified, decisions, next steps) outperforms naive approaches by forcing the summarizer to populate explicit categories. The weakest link across all methods is artifact trail integrity, scoring 2.2-2.5 out of 5.0, so expect to build separate file tracking if your agent needs reliable modification history. Includes probe-based evaluation strategies and a three-phase workflow for compressing large codebases into executable specs.

skills.sh installs: 1.2K
GitHub stars: 43.1K
last updated: Jul 14, 2026

View on GitHub

Install

$ npx -y skills add sickn33/antigravity-awesome-skills --skill context-compression --agent claude-code

Installs into .claude/skills of the current project.

Files

Context Compression Strategies

When agent sessions generate millions of tokens of conversation history, compression becomes mandatory. The naive approach is aggressive compression to minimize tokens per request. The correct optimization target is tokens per task: total tokens consumed to complete a task, including re-fetching costs when compression loses critical information.