A comprehensive safety suite for AI agents providing model pinning, fallback validation, cache staleness detection, and session hygiene.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install safety-checks
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 safety-checks 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 safety-checks?
The safety-checks skill is a robust reliability layer designed to prevent silent degradation in AI systems. By providing unified runtime verification, it ensures that model versions remain consistent, fallback paths are properly configured, and system caches do not lead to stale or corrupted outputs. This tool is essential for developers using Openclaw Skills who require high-assurance environments where model drift or session contamination could compromise the integrity of automated workflows.
Operating entirely on local metadata and configuration files, this skill maintains a strict privacy posture. It acts as a pre-flight checklist and a continuous monitoring heartbeat, ensuring the agent runtime environment precisely matches the developer's expectations before and during execution.
safety-checks Use Cases
- Preventing unexpected model version drift that could lead to inconsistent code generation
- Verifying that secondary and tertiary fallback chains are correctly defined in configuration files
- Identifying and clearing stale cache entries to ensure the agent uses the most recent data
- Detecting cross-session state contamination by identifying orphan lock files and temporary artifacts
How safety-checks Works
- The skill triggers during pre-flight checks or via a periodic heartbeat to assess system health.
- It compares the model version string reported by the agent runtime against the pinned version in the configuration.
- It performs a configuration audit to ensure that fallback entries for models and storage are present and valid.
- It analyzes cache file timestamps against a defined Time-To-Live (TTL) to flag stale data.
- it checks for the existence of stale lock files or orphan temp files in the workspace to ensure session hygiene.
- All results and warnings are written to local log files for remediation or automated enforcement.
safety-checks Setup
Install the skill directly using the CLI:
openclaw install leegitw/safety-checks
For full integration with safety enforcement, install the dependency chain:
openclaw install leegitw/constraint-engine leegitw/context-verifier leegitw/failure-memory leegitw/safety-checks
safety-checks Data Schema & Taxonomy
The skill manages its state and logs within the workspace to maintain transparency:
| Path | Description | Access Mode |
|---|---|---|
.openclaw/safety-checks.yaml |
Primary configuration file for pinned models and TTLs | Read |
output/safety/model-checks.log |
History of model version verification and drift events | Write |
output/safety/session-state.log |
Records of session hygiene checks and lock file status | Write |
.openclaw/cache/staleness.log |
Log of cache age verification and cleanup actions | Read/Write |
safety-checks Advanced Features
- Strict enforcement mode that fails the agent session immediately upon detecting model version drift
- Automated session remediation to clear orphan lock files and temporary data with a single command
- Deep integration with the HEARTBEAT protocol for continuous background monitoring of environment health
- Support for multi-layered configuration, compatible with both Openclaw Skills standards and Claude Code setups
SKILL.md
Loading
Related Openclaw Skills
Patent Validator
An expert research assistant that generates comprehensive search strategies and multi-source queries to help users validate technical concepts against existing implementations.
Patent Scanner
A technical analysis tool that identifies distinctive patterns in concepts using a structured patent attorney methodology.
Moltbook Search
Perform advanced hybrid semantic search across a massive database of over 125,000 AI agent posts from the Moltbook social network.
Memory Garden
A local-first knowledge layer for AI agents that validates patterns through repeated successful use to ensure high-reliability memory.
side-quests
side-quests is a creative synthesis engine that transforms technical insights into a Suno-ready song, a visual concept guide, and a full-length TED talk script.
song-remix
A creative AI agent skill that transforms existing songs into two distinct versions—Respectful and Viral—complete with Suno v4.5 production sliders.
METADATA
Featured*
