A high-reliability continuity kernel providing fail-open context injection, deterministic state persistence, and shadow-mode evaluation receipts.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install continuity-kernel
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 continuity-kernel 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 continuity-kernel?
The continuity-kernel is a specialized framework designed to handle the complexities of state and context management within autonomous agent environments. It enables seamless injection of bounded continuity context into llm_input, ensuring that agents retain critical session data without requiring manual configuration for every turn. By leveraging this tool within the ecosystem of Openclaw Skills, developers can build more resilient systems that survive storage or logic errors through a fail-open design.
Beyond simple persistence, the skill provides a rigorous evaluation layer. It allows developers to generate deterministic runtime contract proofs and execute shadow-mode evals that measure how agents handle perturbations and compaction. This makes it an essential component for production-grade agentic workflows where auditability and reproducible behavior are mandatory.
continuity-kernel Use Cases
- Injecting bounded continuity context into llm_input without manual setup.
- Persisting Soul Card and Mission Ticket states using zero-config SQLite storage.
- Maintaining agent functionality through fail-open logic during continuity errors.
- Generating verifiable runtime proof receipts for compliance and technical audits.
- Running automated shadow-mode evaluations to benchmark agent robustness against noise and data deletion.
How continuity-kernel Works
- The kernel initializes a durable SQLite database (defaulting to ~/.local/state/continuity-kernel/continuity.db) to store persistent state.
- The ContinuityHookAdapter monitors the agent lifecycle, automatically preparing and injecting context into the input stream for the LLM.
- If the storage layer or continuity logic fails, the system triggers a fail-open state, allowing the agent to continue processing rather than crashing.
- Developers use the provided shadow eval scripts to run large-scale simulations (e.g., 100+ runs) against specific Soul Card chunks.
- Evaluation outputs are captured as JSON artifacts, providing a deterministic snapshot of performance and quality metrics.
continuity-kernel Setup
The continuity-kernel requires Python 3. To set up the environment and verify the installation for use with Openclaw Skills, follow these steps:
# Verify Python 3 installation
python3 --version
# Run the core test suite to ensure kernel integrity
python3 -m unittest discover -s tests/continuity-kernel -p 'test_*.py'
# Optional: Set a custom database path for state persistence
export CONTINUITY_KERNEL_DB_PATH="/path/to/custom/storage.db"
# Generate a runtime contract proof to verify the deterministic environment
python3 skills/continuity-kernel/generate_runtime_contract_proof.py
continuity-kernel Data Schema & Taxonomy
The skill organizes data into structured persistence layers and evaluation artifacts to support the Openclaw Skills architecture:
| Data Type | Format | Description |
|---|---|---|
| Persistence Store | SQLite (.db) | Stores Soul Card and Mission Ticket state data at the specified system path. |
| Shadow Traces | JSONL | Logs observed shadow outcomes used for batch evaluation runs. |
| Quality Snapshots | JSON | Contains chunk-level eval receipts and accuracy metrics stored in the artifacts directory. |
| Runtime Proofs | File Artifact | Deterministic proof of the runtime contract for audit logs. |
continuity-kernel Advanced Features
- Multi-suite shadow evaluations supporting core-shadow, memoryarena-mini, and perturbation-based testing.
- Deterministic perturbation profiles including deletion, noise injection, and reordering to test agent robustness.
- Dual-route experimental selector modes for advanced context routing logic.
- Attention-preserving compaction policies that intelligently manage context window limits.
- Configurable artifact root paths for centralized logging and receipt management within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Context Engineer
A powerful context window optimizer designed to analyze, audit, and compress AI agent token utilization.
CacheForge
CacheForge is a token optimization gateway that streamlines onboarding and operations for OpenAI-compatible agent workflows.
vibe-check
An automated auditing tool that identifies unreviewed AI-generated code patterns and produces a scored quality report card.
CacheForge Stats
A terminal-based dashboard for monitoring CacheForge LLM usage, token savings, and performance metrics.
Feed Diet
Analyze and optimize your Hacker News and RSS consumption with AI-generated reports and personalized digests.
kube-medic
A powerful Kubernetes diagnostic toolkit for instant AI-driven cluster triage and incident response using standard kubectl commands.
METADATA
Featured*
