A security analysis tool that identifies dangerous emergent behaviors created by the combination of individually safe agent skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install capability-composition-analyzer
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 capability-composition-analyzer 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 Capability Composition Analyzer?
The Capability Composition Analyzer is designed to solve the capability composition problem, where an agent's total attack surface is greater than the sum of its parts. While individual Openclaw Skills may be benign when audited in isolation, their combination can create critical security vulnerabilities. For instance, a skill that reads local files and another that sends HTTP requests can be combined to exfiltrate sensitive data, even if neither skill is malicious on its own.
This analyzer provides a holistic view of agent security by synthesizing the core value of skill interactions. It moves beyond simple permission checks to model the effective capabilities of an agent. By using this tool, developers can ensure that their Openclaw Skills ecosystems remain secure against emergent risks and inheritance amplification, where a single compromised skill could otherwise inherit the entire agent's power.
Capability Composition Analyzer Use Cases
- Auditing a multi-skill agent to identify hidden exfiltration paths before production deployment.
- Evaluating the security impact of adding a new skill to an existing suite of Openclaw Skills.
- Detecting gaps between an agent's declared permissions and its actual effective capabilities.
- Assessing the potential blast radius and inheritance amplification if a specific skill slot is compromised.
- Monitoring rapid composition changes to prevent the introduction of novel dangerous combinations.
How Capability Composition Analyzer Works
- The user provides the analyzer with a list of an agent's Openclaw Skills, capability metadata, and declared permissions.
- The tool scans for known dangerous pairs, such as combining file system access with network outbound capabilities.
- It identifies the emergent capability surface, documenting functions the agent can perform that no individual skill explicitly declared.
- The analyzer calculates an inheritance amplification score to determine what a poisoned skill would inherit from the existing composition.
- A detailed composition risk report is generated, classifying the agent's safety level from SAFE to CRITICAL and providing remediation advice.
Capability Composition Analyzer Setup
To use the Capability Composition Analyzer, ensure your environment meets the base requirements of curl and python3.
# Verify dependencies
python3 --version
curl --version
Once dependencies are confirmed, integrate the analyzer into your workflow by providing your skill manifest. Ensure all Openclaw Skills in your composition have accurate capability metadata for the most effective analysis.
Capability Composition Analyzer Data Schema & Taxonomy
The analyzer organizes its findings into a structured report to clarify the relationship between individual skills and total agent power:
| Data Point | Description |
|---|---|
| Dangerous Pairs | Inventory of specific skill combinations that create high-risk paths. |
| Emergent Surface | List of undeclared effective capabilities (e.g., secret exfiltration). |
| Inheritance Score | Measure of the potential blast radius for each skill slot. |
| Permission Gaps | Identifies where effective capability exceeds declared scoped permissions. |
| Risk Level | Final categorical assessment: SAFE, ELEVATED, HIGH, or CRITICAL. |
Capability Composition Analyzer Advanced Features
- Structural novelty detection that identifies dangerous compositions even if they do not match known patterns.
- Automated remediation recommendations, such as capability isolation and network allowlisting.
- Deep integration with other security probes for runtime attestation and blast radius estimation.
- Composition change velocity tracking to alert developers when rapid skill updates increase the attack surface.
- Support for complex Openclaw Skills architectures involving multi-agent interactions and shared resources.
SKILL.md
Loading
Related Openclaw Skills
Blast Radius Estimator
A powerful security utility designed to map and project the potential damage of a malicious update within an agent skill dependency graph.
Behavioral Invariant Monitor
A security monitoring tool that detects AI agent skills that maintain safe behavior during initial audits but shift to malicious patterns after multiple runs or specific triggers.
Attestation Root Diversity Analyzer
A diagnostic tool for Openclaw Skills that measures trust root concentration to prevent structural fragility in agent attestation graphs.
Attestation Chain Auditor
A diagnostic tool for verifying the completeness, integrity, and validity of trust attestation chains within AI agent ecosystems.
Capability Graph Mapper
A security analysis tool that uncovers hidden, emergent capabilities created when multiple AI agent skills are combined into a single agent.
Capability Scope Expansion Watcher
A security auditing tool that identifies AI skills gradually claiming broader permissions through subtle, incremental version updates.
Featured*
