session-recovery

What This Skill Does

The session-recovery skill is a sophisticated AI-driven framework designed to solve the critical challenges of long-term AI interaction: context window saturation, high latency, and loss of historical decision-making data. By implementing a robust three-tier documentation system, this skill enables AI agents to maintain persistent states across fragmented conversations. It functions by combining a top-level 'STATUS.md' dashboard for quick status checks, a 'QUICK_RECOVERY.md' index for structural context, and a daily detailed log system ('memory/') for granular audit trails. Beyond documentation, the skill provides a suite of automation scripts that handle the administrative overhead of maintaining these records, including intelligent context compression that can reduce your token usage by up to 90%, significantly accelerating AI response times while lowering costs.

Installation

You can install this skill directly through your agent terminal by running the following command:

clawhub install openclaw/skills/skills/blqbzf/session-recovery

Ensure you have the necessary write permissions in your project directory to allow the scripts to generate and update the memory and status files.

Use Cases

• Long-term Software Development: Managing complex coding projects that span multiple weeks or months where context continuity is essential for bug tracking and feature development.
• Multi-tasking Orchestration: Juggling parallel project threads where the AI needs to swap between different task contexts without forgetting constraints or current progress.
• High-Latency Environments: Projects requiring cost-efficiency and faster token processing, where maintaining a massive, uncompressed chat history is no longer feasible.
• Team-based AI Collaboration: Using the shared memory files as a 'source of truth' that multiple team members or different AI instances can reference to stay aligned on project goals.

Example Prompts

1. "Perform a session recovery: scan the memory folder and update my STATUS.md with today's completed tasks."
2. "I'm hitting a 15k token limit; trigger the context compression strategy and summarize our progress into a new summary index."
3. "Initialize a new daily log for our project and include the standard template for task tracking and decision logging."

Tips & Limitations

• Consistency is Key: The system is most effective when you run the update_status.sh script at the end of every work cycle. If the documentation becomes stale, the recovery process will be inaccurate.
• Compression Thresholds: While the skill allows for aggressive compression, ensure you keep at least one full 'memory' file for historical auditing before performing a total archival sweep.
• Human-in-the-loop: The AI manages the documentation, but human review is recommended for the 'Key Decisions' section in the QUICK_RECOVERY.md file to ensure the AI hasn't hallucinated or misinterpreted complex logic.