Performs a hard reset for specific task types by purging learned playbooks, confidence data, and evaluation history.
ALM Forget is a vital maintenance utility for the Claude Code Learning Memory (ALM) system, allowing users to selectively remove learned data associated with specific task types. Whether a workflow has changed, a framework has been deprecated, or an experimental task type is no longer needed, this skill ensures the agent's memory remains accurate by deleting associated playbooks, confidence scores, taxonomy entries, and evaluation records while resetting the underlying classifier model.
Key Features
01Interactive discovery mode to list all currently known task types
02Targeted deletion of playbooks and confidence entries
03Reflect queue management to prevent processing of deleted tasks
04Automated cleanup of taxonomy and evaluation record files
05TF-IDF classifier model reset to ensure memory integrity
060 GitHub stars
Use Cases
01Purging outdated coding standards or best practices after a project migration
02Cleaning up the ALM directory by removing unused or experimental task classifications
03Resetting the learning state for a specific task to retrain the model from scratch
