self-improving-agent-3 Security Audit Report

🔭 Continuously monitored by ClawSecure Watchtower

Source:

SHA-256:

self-improving-agent-3 is an AI agent skill. ClawSecure audited self-improving-agent-3 across 41 files through the 3-Layer Audit Protocol covering all ten OWASP ASI Top 10 categories, assigning a security score of 30/100 (High Risk). The 8 findings concentrate in Malicious Code, Permissions Manifest and ReDoS, including Attempts to access sensitive file: SOUL.md and Attempts to access sensitive file: MEMORY.md. 6 were rated high or critical severity.

Is self-improving-agent-3 safe?

ClawSecure audited self-improving-agent-3 and assigned a security score of 30/100 (High Risk), identifying 8 findings across Malicious Code and Permissions Manifest. Review the findings below before installing.

What did ClawSecure find in self-improving-agent-3?

ClawSecure identified 8 findings in self-improving-agent-3, concentrated in Malicious Code, Permissions Manifest and ReDoS. 6 were rated high or critical severity. The most severe include Attempts to access sensitive file: SOUL.md and Attempts to access sensitive file: MEMORY.md.

How was self-improving-agent-3 audited?

ClawSecure ran self-improving-agent-3 through its 3-Layer Audit Protocol with full OWASP ASI Top 10 coverage, scanning 41 files.

What does a score of 30 mean?

ClawSecure assigned self-improving-agent-3 a security score of 30/100, placing it in the High Risk range. This is driven by 8 findings led by Malicious Code that should be addressed before use. ClawSecure derives this score with a weighted deduction model (critical -20, high -10, medium -5, low -2 from a base of 100).

Audit Findings for self-improving-agent-3

ClawSecure detected 8 security findings in self-improving-agent-3, spanning Malicious Code, Permissions Manifest and ReDoS.

high · Attempts to access sensitive file: SOUL.md. Malicious Code finding detected in SKILL.md.
high · Attempts to access sensitive file: MEMORY.md. Malicious Code finding detected in SKILL.md.
high · Attempts to access sensitive file: SOUL.md. Malicious Code finding detected in handler.js.
high · Attempts to access sensitive file: SOUL.md. Malicious Code finding detected in handler.ts.
high · Attempts to access sensitive file: SOUL.md. Malicious Code finding detected in openclaw-integration.md.
high · Attempts to access sensitive file: MEMORY.md. Malicious Code finding detected in openclaw-integration.md.
medium · Missing config.json - agent may not be properly configured. Permissions Manifest finding.
medium · ReDoS vulnerability: Nested quantifiers (potential…. ReDoS finding detected in extract-skill.sh:101.

Each finding is expandable in the interactive list below.

3-Layer Audit Protocol

Security Recommendations for self-improving-agent-3

Audit external network connections

self-improving-agent-3 connects to external endpoints. Verify every outbound connection goes to a trusted destination. Unauthorized callbacks are a primary indicator of ClawHavoc malware and data exfiltration. ClawSecure's proprietary engine monitors for known malicious endpoints including C2 infrastructure.

Add a config.json permissions manifest

A config.json file declares what an agent component can access: file system, network, shell execution and more. Without it, users have no visibility into what the component can do before installing. This is the single most impactful security improvement for any AI agent skill.

Fix ReDoS-prone patterns

self-improving-agent-3 contains regular expressions vulnerable to catastrophic backtracking (ReDoS). Replace vulnerable patterns, bound input length, and prefer linear-time matching so a crafted input cannot hang the agent.

Related Security Research

ClawHavoc Explained: The Malware Family Targeting AI Agents→Beyond Static Scans: Why ClawSecure Verifies Agentic Intent→

Related AI Agent Security Audits

mission-controlScore 38/100 @openclaw/dashboardScore 45/100 cron-creatorScore 35/100 claw-controlScore 25/100 jasper-recallScore 15/100

Scanned on March 17, 2026. self-improving-agent-3 is one of thousands of agents audited by ClawSecure from the community-curated awesome-openclaw-skills list and the openclaw/skills repository.

Start Your Free Scan