A persistent memory management skill that enables AI agents to store, search, and update long-term task and project data using the Milvus vector database.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install long-term-task-memory
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install long-term-task-memory using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Long-term Task Memory?
Long-term Task Memory provides AI coding agents with the ability to maintain context across long intervals by leveraging the Milvus vector database. As a key component of Openclaw Skills, it allows agents to store detailed records of tasks, events, and project decisions that persist far beyond a single conversation session. This ensures that the agent remains context-aware of historical technical choices and project evolutions.
The skill is built for flexibility, supporting multi-dimensional data categorization. Whether you are tracking a developer's specific task progress or a manager's high-level architectural decisions, this skill organizes information into a searchable, structured format. By using Openclaw Skills like this one, developers can build more reliable and intelligent autonomous workflows that don't lose track of critical project history.
Long-term Task Memory Use Cases
- Archiving complex technical decisions and architectural patterns for long-term project continuity.
- Tracking task statuses and sprint progress across multiple AI-driven development sessions.
- Managing role-based project memories to provide different contexts for PMs, developers, and designers.
- Building a searchable knowledge base of meeting outcomes and event-driven project updates.
How Long-term Task Memory Works
- The skill establishes a secure connection to a Milvus database instance using URI and Token credentials provided in the environment configuration.
- Information is captured in a structured JSON format, including content, categories, roles, and project tags.
- The data is vectorized and stored in a specified collection (defaulting to task_memory), making it available for semantic and filtered search.
- Users or agents can query the memory using specific metadata filters such as project name, status, or priority level.
- The lifecycle of a memory is managed through update and delete actions, allowing the AI to refine or remove information as the project evolves.
Long-term Task Memory Setup
To integrate this capability into your workflow, ensure you have the necessary dependencies and environment variables configured. This is a core part of the Openclaw Skills ecosystem.
# Install the required Milvus client
pip install pymilvus==2.3.0
# Configure your .env file with Milvus instance details
MILVUS_URI=http://your-instance.milvus.ivolces.com:19530
MILVUS_TOKEN=root:yourpassword
# Initialize the database collection
python scripts/milvus_manager.py --action init
Long-term Task Memory Data Schema & Taxonomy
The skill utilizes a flexible schema to handle diverse data types. Below is the primary structure used for organizing information within the Openclaw Skills framework:
| Dimension | Field | Description | Example |
|---|---|---|---|
| Category | category | The type of record | task, event, project, decision |
| Role | role | The persona involved | pm, dev, manager, designer |
| Project | project | Project identifier | website-redesign, app-v2 |
| Status | status | Current lifecycle state | pending, in_progress, completed |
| Priority | priority | Importance level | high, medium, low |
| Tags | tags | Custom identifiers | ["urgent", "backend", "api"] |
| Context | context | Custom JSON data | { "tech_stack": ["FastAPI"] } |
Long-term Task Memory Advanced Features
- Multi-collection support allowing for segregated memory silos for different organizations or massive projects.
- Full CRUD operations via CLI for manual memory management and data cleanup.
- Flexible JSON metadata and context support for storing arbitrary technical data points.
- Semantic search combined with boolean filtering for high-precision information retrieval.
- Automatic environment configuration loading from multiple directory levels for easy deployment.
SKILL.md
Loading
Related Openclaw Skills
Image to SVG
A utility that converts common raster image formats into SVG containers with options for base64 embedding or local linking.
Filewatch Reference Skill
An offline reference tool for Filewatch concepts, best practices, and implementation patterns designed for rapid developer lookup.
Logrotate Reference Skill
A local, high-performance reference tool providing instant access to Logrotate implementation patterns, security best practices, and debugging guides.
Mailcheck Reference
A lightweight reference tool providing instant access to Mailcheck implementation patterns, debugging guides, and performance best practices.
Linecount Reference Skill
A local, high-performance reference tool providing instant access to Linecount development concepts and implementation patterns.
Beat
A local-first command-line tool for tracking, analyzing, and managing music and audio files with full data privacy.
Featured*
