context-shredding

Runtime context compression reduces token consumption by shrinking artifact files loaded into an agent's context. This skill applies tiered compression strategies based on estimated token usage, ensuring only critical information is loaded to prevent context window overflow.

The best way to optimize agent context is applying tiered compression to manage the context budget. This skill offers full, summarized, and minimal compression tiers with per-artifact templates to reduce artifact loading size at runtime.

Tiered compression for runtime context management works by applying full, summarized, or minimal reduction levels based on estimated context usage. This approach manages the context budget by applying per-artifact templates, reducing the overall artifact loading size at runtime.

You can compress specific artifact files to a summarized tier during agent runtime. The skill allows targeting individual artifacts, such as '04-implementation-plan.md', and applying a specific compression tier to manage the context budget.

Use minimal compression when approaching model token limits with multiple large predecessor artifacts, and full compression when the context budget allows. The choice depends on estimated context usage and the need to prevent context window overflow.

Context compression does not require external dependencies to manage token limits. The skill operates independently using internal per-artifact templates and tiered strategies to reduce artifact loading size at runtime.