Summarize conversation context to preserve goals, progress, and key decisions.
npx skills add https://github.com/sinaptia/detritus --skill compact-sinaptia
Please help me install this Agent Skill. Skill: compact Source: https://github.com/sinaptia/detritus/tree/main/skills/compact Command: npx skills add https://github.com/sinaptia/detritus --skill compact-sinaptia
This Skill addresses the challenge of lengthy conversation histories by creating concise, structured summaries that preserve essential information while freeing up valuable context window space.
Use the compact skill to create a summary of the current conversation.
High-intent search queries and answers about installing and using this skill.
To preserve agent state in long-running interactions, it captures key decisions, progress, and critical context in a structured summary format, allowing the AI to continue efficiently without losing track of the project's state.
You need to compact a lengthy conversation history when a complex coding session becomes long, requiring you to archive older messages and summarize progress, decisions made, and current blockers to manage context window length.
The best way to preserve key decisions during long-running agent interactions is to apply context compaction, which captures progress and critical context in a structured format to maintain state preservation for continued efficient operation.
Context compaction does not lose critical project state information, as it specifically preserves essential information including goals, progress, key decisions, and critical context while reducing the amount of text the AI needs to process.
You can manage context window length for complex coding sessions without losing track by archiving older messages into a structured summary, ensuring the AI continues efficiently with the project's state intact.
EliottMeunierFluidGuide structured conversation and data capture to initialize an AI-aware vault.
sethihqMonitor, compact, and resume AI context windows via handoffs.
kouisoSuggest manual context compaction at logical intervals during multi-phase tasks.
daehyunpySuggest manual context compaction at configurable tool-use thresholds.
