keep is an AI agent skill, created by hughpyle and published at openclaw/skills. ClawSecure audited keep across 74 files through the 3-Layer Audit Protocol covering all ten OWASP ASI Top 10 categories, assigning a security score of 43/100 (Warning). The 11 findings concentrate in Data Exfiltration, Malicious Code and Unauthorized Tool Use, including Attempts to access sensitive file: MEMORY.md and Attempts to access sensitive file: MEMORY.md. 2 were rated high or critical severity.
Is keep safe?
ClawSecure audited keep and assigned a security score of 43/100 (Warning), identifying 11 findings across Data Exfiltration and Malicious Code. Review the findings below before installing.
What did ClawSecure find in keep?
ClawSecure identified 11 findings in keep, concentrated in Data Exfiltration, Malicious Code and Unauthorized Tool Use. 2 were rated high or critical severity. The most severe include Attempts to access sensitive file: MEMORY.md and Attempts to access sensitive file: MEMORY.md.
How was keep audited?
ClawSecure ran keep through its 3-Layer Audit Protocol with full OWASP ASI Top 10 coverage, scanning 74 files from openclaw/skills.
What does a score of 43 mean?
ClawSecure assigned keep a security score of 43/100, placing it in the Warning range. This is driven by 11 findings led by Data Exfiltration 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 keep
ClawSecure detected 11 security findings in keep, spanning Data Exfiltration, Malicious Code, Unauthorized Tool Use and Social Engineering.
- high · Attempts to access sensitive file: MEMORY.md. Malicious Code finding detected in
KEEP-FIND.md. - high · Attempts to access sensitive file: MEMORY.md. Malicious Code finding detected in
OPENCLAW-INTEGRATION.md. - medium · Skill code uses network libraries but doesn't declare network.... Unauthorized Tool Use finding detected in
/tmp/url-scans/728a88eb93672bc6/SKILL.md. - medium · Script iterates through environment variables in.... Data Exfiltration finding detected in
/tmp/url-scans/728a88eb93672bc6/bench/locomo/llm.py. - medium · Pattern detected: urllib.request.Request(. Data Exfiltration finding detected in
bench/locomo/ingest.py:215. - medium · Pattern detected: urllib.request.urlopen(. Data Exfiltration finding detected in
bench/locomo/ingest.py:220. - medium · Variable $CURSOR (line 9) flows to `sh` at line 16.. Command Injection finding detected in
docs/KEEP-FLOW.md:123. - medium · File 'han_verse.txt' extension (.txt) suggests one format but Magika.... Obfuscation finding detected in
docs/library/han_verse.txt. - medium · Missing config.json - agent may not be properly configured. Permissions Manifest finding.
- low · Skill description is too short (17 chars). Provide detailed explanation.. Social Engineering finding detected in
SKILL.md. - info · Skill manifest does not include a 'license' field. Specifying a license.... Policy Violation finding detected in
SKILL.md.
Each finding is expandable in the interactive list below.
3-Layer Audit Protocol
Security Recommendations for keep
Audit external network connections
Resolve policy violations
Harden command execution
Related Security Research
ClawHavoc Explained: The Malware Family Targeting AI Agents→Beyond Static Scans: Why ClawSecure Verifies Agentic Intent→Related AI Agent Security Audits
Scanned on April 30, 2026. keep is one of thousands of agents audited by ClawSecure from the community-curated awesome-openclaw-skills list and the openclaw/skills repository.