learn-reset
Scannednpx machina-cli add skill aiskillstore/marketplace/learn-reset --openclawLearn Reset
Clear all accumulated knowledge and reset to a fresh state.
What This Does
- Clears all entries from
knowledge/learnings/files (patterns, quirks, decisions) - Resets the classification cache
- Resets learning state (extraction count, queries)
- Preserves file structure (doesn't delete files)
Warning: This action cannot be undone. All accumulated insights will be lost.
Instructions
- Confirm with user - This is destructive, ask for confirmation first
- Reset learnings files - Clear entries from:
knowledge/learnings/patterns.mdknowledge/learnings/quirks.mdknowledge/learnings/decisions.md
- Reset cache - Clear
knowledge/cache/classifications.md - Reset session - Clear
knowledge/context/session.md - Reset state - Reset
knowledge/state.jsonto initial values - Confirm completion
Reset File Format
After reset, each learnings file should have:
---
type: [type]
version: "1.0"
description: [original description]
last_updated: null
entry_count: 0
---
# [Title]
[Description]
**Purpose:** [Purpose]
---
<!-- Entries will be appended below this line -->
State Reset
Reset knowledge/state.json to:
{
"version": "1.0",
"learning_mode": false,
"learning_mode_since": null,
"last_extraction": null,
"extraction_count": 0,
"queries_since_extraction": 0,
"extraction_threshold_queries": 10,
"extraction_threshold_minutes": 30
}
Output Format
Knowledge Base Reset
────────────────────
Are you sure you want to clear all knowledge? This cannot be undone.
[After confirmation]
Knowledge base has been reset:
- Cleared 8 patterns
- Cleared 3 quirks
- Cleared 5 decisions
- Cleared 23 cached classifications
- Reset learning state
The knowledge base is now empty. Use /learn to start fresh.
Notes
- Always confirm before resetting
- This does not delete the knowledge directory structure
- Learning mode is disabled after reset
- Git history may still contain old knowledge if previously committed
Source
git clone https://github.com/aiskillstore/marketplace/blob/main/skills/0xrdan/learn-reset/SKILL.mdView on GitHub Overview
Learn-reset fully clears accumulated knowledge and resets learning state to a fresh baseline. It empties the main learnings files (patterns, quirks, decisions), clears the classification cache, and resets session and state, while preserving the directory structure. This action is destructive and cannot be undone, so a user confirmation is required.
How This Skill Works
The skill deletes entries from knowledge/learnings/patterns.md, knowledge/learnings/quirks.md, and knowledge/learnings/decisions.md; it clears knowledge/cache/classifications.md; it resets knowledge/context/session.md and resets knowledge/state.json to the initial values. After completion, learning mode is disabled and the system is ready to collect new insights, with the original file structure intact.
When to Use It
- Starting a brand-new domain or project to avoid legacy learnings.
- Correcting corrupted or unwanted learned entries that skew behavior.
- Preparing a sandbox or test environment for fresh evaluation.
- Reproducing a clean baseline before a major knowledge refresh.
- Diagnosing persistent misclassifications after repeated learning drift.
Quick Start
- Step 1: Initiate Learn Reset via the appropriate admin or user command.
- Step 2: When prompted, confirm the destructive reset to proceed.
- Step 3: Review the completion message and verify all targeted files and state have been reset.
Best Practices
- Always prompt for explicit user confirmation before triggering the destructive reset.
- Back up any valuable learnings or notes before proceeding.
- Verify that all learnings files are cleared and show zero entries after reset.
- Confirm that knowledge/cache and state have been reset to initial values.
- Disable learning mode after reset to prevent automatic re-learning during setup.
Example Use Cases
- A team reintroduces domain rules from scratch after a tooling upgrade.
- A bot migrates to a new knowledge schema and performs a full reset for a clean baseline.
- QA runs a reset to validate the reset procedure and ensure a pristine environment.
- Data drift evaluation is concluded and the knowledge base is cleared for fresh testing.
- A sandbox environment is reset prior to Live testing to avoid cross-environment contamination.
