skills.rest
Helcio-Nogueira
Helcio-NogueiraCOMMUNITY
Explore all skills by this author →
GitHub

context-compression

Summarizes completed work phases to compress long conversation contexts.

Updated Mar 23, 2026
Search Tags:#summarization#context management#ai agent#compression#token optimization#long sessions#session checkpoint
One-click install
npx skills add https://github.com/Helcio-Nogueira/Painel_Inteligente_Hackaton_2026 --skill context-compression-helcio-nogueira
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-compression
Source: https://github.com/Helcio-Nogueira/Painel_Inteligente_Hackaton_2026/tree/main/.agents/skills/context-compression
Command: npx skills add https://github.com/Helcio-Nogueira/Painel_Inteligente_Hackaton_2026 --skill context-compression-helcio-nogueira

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long AI sessions (20+ turns) cause context degradation: the agent repeats itself, forgets earlier decisions, and loses track of completed work. This Skill detects context saturation and compresses finished phases into concise summaries while preserving key decisions. ## Core Features & Use Cases - Three Compression Levels: Micro-compact large tool outputs, summarize completed work phases, and create full session checkpoints with decisions and file references. - Compression Protocol: A three-step process to identify completed phases, extract key decisions and file paths, and write 100-300 token summaries. - Use Case: During a 35-turn debugging session, compress the completed research phase into a 200-token summary listing the auth flow, bug location at jwt.ts:45, and the fix decision, freeing thousands of tokens for the implementation phase. ## Quick Start Ask the agent to summarize what has been done so far and compress the completed phases of this session.

Frequently Asked Questions about context-compression

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I compress context in a long AI conversation?▼

Identify completed work phases, extract key decisions and file references, then replace verbose transcripts with 100-300 token summaries. Use micro-compaction for large tool outputs and session checkpoints at phase transitions.

When should I summarize conversation history?▼

Summarize when a session exceeds 20 turns, when the agent repeats earlier suggestions, when the user says something was already discussed, or when switching between work phases like research to implementation.

What information should be preserved when compressing context?▼

Preserve decisions made and their rationale, file paths with line numbers, key findings that inform ongoing work, and summarized error messages. Discard step-by-step tool invocations, full file contents, and exploratory dead-ends.

How many tokens can context compression recover?▼

Context compression typically recovers 5,000 to 15,000 tokens in long sessions by replacing verbose tool outputs and research transcripts with semantic summaries of 100-300 tokens per completed phase.

What are the limitations of context compression?▼

Compression should never be fully automatic without notifying the user, and it risks losing detail if summaries omit file references or decision rationale. It is not useful for short sessions under 20 turns.

Related Skills

Explore More Software Engineering Skills→
arbisoftarbisoft
4

blueprint

Turn a single objective into a modular multi-session construction plan.

Official
Advanced
TengHuTengHu

pod-resume-state

Load the most recent saved checkpoint across theses or within a specified thesis.

Community
Advanced
vudovnvudovn
8.2k

context-compression

Compresses long conversation sessions into summaries that preserve key decisions and file references.

Community
Intermediate
Hidden-HistoryHidden-History
41

parzival-save-handoff

Saves Parzival session handoff documents to the Qdrant discussions collection.

Community
Intermediate