Compact session

Skill couzip/cc-forever/skills/compact-session

Give Claude Code a persistent memory. Save and search past conversations with semantic search. 100% local, no API keys required.

Install

npx -y skills add couzip/cc-forever --skill compact-session

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Summarize entire session and save to index as compact Q&A pairs

SKILL.md

1.3 KB, as published. Nobody here has run it

Compact Session

CRITICAL RULE

store_memory must be called EXACTLY ONCE. No retries. No corrections.

Pre-call Checklist (ALL must be YES before calling store_memory)

Identified ALL main topics (3-5)?
Written summaries for EVERY topic?
Combined ALL summaries into ONE content string?
Displayed the complete content string?

Process

List all main topics discussed in this session (3-5 topics)
For each topic, write: "Human: [question]\nAssistant: [answer]"
Concatenate all pairs with blank lines between them
Display the COMPLETE content string
Call store_memory ONCE with that complete string

Format Example

Content string should look like:

Human: How to setup CC-Forever?
Assistant: Run /cc-forever:setup to configure.

Human: What embedding models are available?
Assistant: English (MiniLM), Multilingual (e5), Japanese (Ruri).

store_memory Call

content: The COMPLETE string containing ALL topic summaries
project: current directory name
tags: ["compact", "session-summary"]

After Calling

Report: "Saved N topic summaries" and STOP. Do not call store_memory again.