checkpoint

Saves conversation context into project files so a new chat can continue work.

Updated Nov 21, 2025
One-click install
npx skills add https://github.com/alatyshau/duet --skill checkpoint-alatyshau
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: checkpoint
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/tools/checkpoint
Command: npx skills add https://github.com/alatyshau/duet --skill checkpoint-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long AI conversations accumulate decisions, open questions, and artifacts that vanish when the chat ends or context is compressed. This Skill packs the conversation into project files so a new chat resumes work without repeating covered ground. ## Core Features & Use Cases - Conversation pass table: Walks every message, classifies each substantive point as already saved or needing capture, and presents the table for user confirmation before writing. - Structured saving: Updates project artifacts, relevant documentation, and target artifacts, marking the current step with [ACTIVE] in the entry-point file. - Verification and handoff: Re-reads saved files for completeness and coherence, then produces an initiating phrase for the new chat plus optional instruction-improvement signals. - Use Case: After a long design session on a product spec, trigger the checkpoint to capture all decisions and open questions into plan.md, then start a fresh chat that picks up exactly where you stopped. ## Quick Start Say "!упакуй" or ask the assistant to checkpoint the conversation so a new chat can continue the project.

Frequently Asked Questions about checkpoint

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

FAQPage Schema
How do I save chat context so a new conversation can continue?

Trigger the checkpoint workflow, which walks the conversation message by message, builds a pass table of substantive points, and saves everything not already in files. After your confirmation, it updates project artifacts and marks the current step with [ACTIVE].

What files does the checkpoint process update?

It updates three categories: project artifacts (plan.md, topic files), relevant documentation (README, PRODUCT.md, COMPONENT.md), and target artifacts (code, instructions, specs). The exact files depend on the project structure determined during the walkthrough.

Does checkpoint work when the conversation was compressed?

It flags a compression risk section in the output when the beginning of a long conversation may have been lost. If important early decisions were not saved along the way, the report notes what might be missing so you can reconstruct it.

Can I review what gets saved before files are written?

Yes. The workflow stops after presenting the conversation pass table and packing plan, and only proceeds to saving after your explicit confirmation. You can edit the plan and have it re-presented before any writes occur.

What are the limitations of conversation checkpointing?

It relies on the agent's judgment of what is substantive, so ephemeral context or implicit reasoning may not be captured. It also requires existing project files to be readable, and cannot recover content already lost to context compression.