A comprehensive configuration and runtime optimization system to manage persona, memory, and security for OpenClaw agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ai-sidekick
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 ai-sidekick 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 AI Sidekick?
The AI Sidekick skill is a technical configuration framework designed to structure the behavior, memory models, and security of agents within your Openclaw Skills toolkit. By isolating persona attributes into SOUL.md and user telemetry into USER.md, the skill enables developers to sculpt an agent's style, response limits, and conversational boundary lines. This architecture facilitates context-aware workflows while keeping operational data secure.
In addition to customization, the skill features automated memory retention mechanics that segment learning into daily logs and long-term project files. Backed by token conservation features and structural sandbox enforcement, AI Sidekick delivers a reliable foundation for running specialized coding agents locally and securely.
AI Sidekick Use Cases
- Creating custom personas with clear boundary parameters and behavioral styles for targeted software engineering tasks.
- Maintaining secure, local user context and habits that persist without leaking details to collaborative environments.
- Implementing prompt injection defenses and sandboxed execution workspaces to secure system directories.
- Organizing project outcomes and technical issues into structured short-term and long-term storage configurations automatically.
How AI Sidekick Works
- System Loading: At startup, the agent initializes by parsing SOUL.md, USER.md, and MEMORY.md from the workspace root to mount context.
- Dynamic Contextualization: During execution, the agent tracks conversational feedback to update preferences, timezone fields, and tooling habits.
- Memory Persistency: On task completion, the skill structures output states, documenting daily changes into the memory directory and consolidating core technical logs.
- Proactive Inspection: High-risk activities, folder operations, and external script interactions trigger sandbox sweeps and permission checks.
- Optimization: Token usage is regulated via segment-based file retrieval alongside structural cleanup of historical daily records.
AI Sidekick Setup
Install the skill into your local workspace:
skillhub_install install_skill ai-sidekick
Define your persona parameters in SOUL.md within your workspace directory:
## Style
Concise and analytical. Reply with direct files and minimize greetings.
Configure the human operator details in USER.md:
# USER.md - About Your Human
- **Name**: Lead Engineer
- **What to call them**: Captain
- **Timezone**: Asia/Shanghai
- **Notes**: Prefers standard shell syntax and minimal explanations.
Restart your gateway controller to load the changes:
gateway restart
AI Sidekick Data Schema & Taxonomy
The skill relies on specific workspace directories and files to map metadata parameters:
| File / Directory | Scope of Use | Operations |
|---|---|---|
SOUL.md |
Core agent persona and communication limits. | Read at startup |
USER.md |
User identity tracking and context parameters. | Updated during execution |
MEMORY.md |
Global project milestones and setup exceptions. | Read at main startup |
memory/ |
Directory for daily journal records (YYYY-MM-DD.md). |
Written post-session |
HEARTBEAT.md |
Interval schedules and diagnostic steps. | Read on heartbeat |
Backup Mechanics
Run the built-in backup shell routine to export memory matrices safely:
# Creates a compressed archive inside ~/.qclaw/backups/
./scripts/backup_memory.sh
AI Sidekick Advanced Features
- Four-Layer Guard: Employs rigorous system checks, command injection parsing, workspace directory locks, and secret masking.
- Automated Memory Rotations: Periodically reviews logs, moving static workflows to system definitions and pruning records older than 90 days to conserve space.
- Dynamic Preference Catching: Observes developer complaints or instructions in conversational loops to append style guidelines instantly.
- Automated Security Scanning: Incorporates diagnostic utilities to inspect imported tools for bad shell patterns, eval blocks, or unauthorized remote connections.
SKILL.md
Loading
Related Openclaw Skills
LipVoice TTS
LipVoice TTS turns reference audio into reusable voice models and generates high-quality WAV speech from text through a simple CLI workflow.
Life Book Generator
The Life Book Generator skill enables AI agents to coordinate the complete creation, payment, and retrieval of 18-chapter life reports via the official Agent API.
Publisher
Publisher lets you publish HTML frontend projects to PushWebly, manage visibility, list projects, and generate Android APKs from the same workflow.
Publisher
Deploy static HTML projects to the cloud and compile them into Android APK packages using automated publishing workflows.
Taiwan Stock Exchange (TWSE) API
A developer integration that enables programmatic retrieval and analysis of Taiwan Stock Exchange open data without authentication.
novel-master
A publication-grade Chinese long-form novel creation framework powered by a unique distributed multi-agent architecture.
Featured*
