Security Audit Report

Score: 43/100
Risk Level: High Risk
Created by: hughpyle
Files Audited: 74 files
Date: 4/30/2026, 1:55:16 PM
Monitored by: ClawSecure Watchtower

Findings:

  • Total Findings: 11
  • Critical: 0
  • High: 2
  • Medium: 7
  • Low: 1
  • Info: 1

Source: https://clawhub.ai/hughpyle/keep
SHA-256: c0477b026bc4...

keep is an Other skill for the OpenClaw ecosystem, created by hughpyle. ClawSecure audited this skill through our 3-Layer Audit Protocol covering all 10 OWASP ASI Top 10 categories. This skill received a security score of 43/100, indicating significant security concerns that require attention. The audit identified 11 findings, with issues detected across multiple security layers.

3-Layer Audit Protocol

🛡 ClawSecure Proprietary Engine

  • Missing config.json - agent may not be properly configured.
  • Attempts to access sensitive file: MEMORY.md

Critical Findings:
3 findings

🔒 Advanced Static & Behavioral Analysis

  • Skill code uses network libraries but doesn't declare network requirement.
  • Script iterates through environment variables in /tmp/url-scans/728a88eb93672bc6/bench/locomo/llm.py.
  • Skill description is too short (17 chars). Provide detailed explanation.
  • 5 more findings.

Warning Findings:
8 findings

📦 Supply Chain Security

  • No issues detected

Pass Findings:
0 findings

HIGH findings

  • F-001 Attempts to access sensitive file: MEMORY.md - ASI06 OWASP ASI: Memory & Context Poisoning
  • F-002 Attempts to access sensitive file: MEMORY.md - ASI06 OWASP ASI: Memory & Context Poisoning

MEDIUM findings

  • F-001 Skill code uses network libraries but doesn't declare network requirement - Advanced Static & Behavioral Analysis
  • F-002 Script iterates through environment variables in /tmp/url-scans/728a88eb93672bc6/bench/locomo/llm.py.
  • F-003 Pattern detected: urllib.request.Request(
  • F-004 Pattern detected: urllib.request.urlopen(
  • F-005 Variable $CURSOR (line 9) flows to `sh` at line 16.
  • F-006 File 'han_verse.txt' extension (.txt) suggests one format but Magika detected a different text format: Markdown document (markdown).
  • F-007 Missing config.json - agent may not be properly configured - ASI03 OWASP ASI: Identity & Privilege Abuse

LOW findings

  • F-001 Skill description is too short (17 chars). Provide detailed explanation.

INFO findings

  • F-001 Skill manifest does not include a 'license' field. Specifying a license helps users understand usage terms.

OWASP ASI Framework Coverage

Covered ASI Categories:

  • ASI01 Agent Goal Hijack
  • ASI02 Tool Misuse
  • ASI03 Identity & Privilege
  • ASI04 Agentic Supply Chain
  • ASI05 Unexpected Code Execution
  • ASI06 Memory & Context
  • ASI07 Insecure Inter-Agent Communication
  • ASI08 Cascading Failures
  • ASI09 Human-Agent Trust Exploitation
  • ASI10 Rogue Agents

Categories highlighted were detected in this scan. ClawSecure provides comprehensive coverage across all 10 OWASP ASI Top 10 categories.

Context-Aware Analysis

This skill requests system-level capabilities which are standard for OpenClaw automation skills. Generic scanners often flag these capabilities as suspicious regardless of context. ClawSecure's Context-Aware Intelligence analyzes capabilities in ecosystem context — we differentiate real threats (malicious code, unauthorized data exfiltration, supply chain vulnerabilities, ReDoS patterns) from normal agent functionality. Our 3-Layer Audit Protocol is purpose-built for the OpenClaw ecosystem.

Security Recommendations for keep

  • Pin npm dependencies to exact versions
    Unpinned dependencies allow supply chain attacks where a compromised package version is automatically pulled into your skill. Use exact version numbers in package.json (e.g., 1.2.3 instead of ^1.2.3) to prevent unauthorized code from entering your dependency tree. ClawSecure's supply chain scanning checks every dependency against known CVE databases.

  • Add a config.json permissions manifest
    A config.json file declares what permissions your OpenClaw skill needs (file system access, network requests, shell execution, etc.). Without it, users have no visibility into what your skill can do before installing. Adding a permissions manifest is the single most impactful security improvement for any OpenClaw skill.

  • Review shell execution permissions
    This skill uses shell execution capabilities. While shell access is standard for many useful OpenClaw automations, it's important to validate that commands are constructed from trusted inputs only. Avoid passing user-controlled strings directly to shell commands, and consider restricting execution to a specific set of allowed commands.