progressive-memory

What This Skill Does

The Progressive Memory skill is a meta-utility designed to optimize the context window for OpenClaw agents. It solves the issue of "context bloat," where a primary memory file (MEMORY.md) becomes too large, causing the agent to lose focus or consume unnecessary tokens. By offloading detailed, non-essential data into separate, indexed markdown files, the agent maintains a lightweight core while retaining the ability to retrieve deep-dive information on-demand. This skill provides two core tools: memorize, which handles the creation of sub-files and automatically updates the central index, and recall, which fetches the specific contents of a memory topic.

Installation

To integrate this into your agent, run the following command in your terminal within the OpenClaw environment: clawhub install openclaw/skills/skills/autogame-17/claw-progressive-memory

Use Cases

This skill is ideal for projects with long-lived state or complex documentation requirements. Common use cases include: 1. Project Specs: Storing detailed API documentation or technical requirements that aren't needed in every interaction. 2. Persona Management: Maintaining complex backstories or visual descriptions that are only relevant when writing specific character narratives. 3. Session Logs: Archiving historical logs or experimental results for future comparison without polluting the active workspace. 4. Preference Lists: Managing large lists of configurations, presets, or user-specific settings that only require attention during initialization or updates.

Example Prompts

1. "Memorize my current project style guide: use the code style defined in the attached documentation and save it to 'style_guide'."
2. "Recall the 'suno_presets' memory file so we can adjust the parameters for the new track."
3. "Store these recent technical findings under 'bug_research_01' using the progressive memory tool to keep my main index clean."

Tips & Limitations

• Tip: Always summarize the content of the sub-file within the MEMORY.md entry so the agent knows exactly when it should trigger a recall.
• Tip: Keep your topic names short and descriptive, such as 'api_keys_config' or 'user_preferences', to make searching easier.
• Limitation: The agent must be capable of deciding when to use the tool. If the agent does not recognize the need for the file, you may need to explicitly prompt it to recall the information.
• Best Practice: Audit your memory/ folder periodically to delete or archive old topics that are no longer relevant to the current project goals.