Scan Report
0 /100
agent-security-hardening
Security hardening patterns for production AI agents. Covers prompt injection defense, data boundary enforcement, read-only defaults, WAL protocol, health check scripts, integrity gates, rule escalation ladder, and session memory security.
This is a legitimate security hardening guide for AI agents containing only documentation with defensive examples. No malicious functionality detected.
Safe to install
This skill is safe to use. No action required.
1 Critical 6 findings
Critical Dangerous Command 危险 Shell 命令
rm -rf / SKILL.md:60 Medium External URL 外部 URL
https://evil.com/exfil SKILL.md:79 Info Email 邮箱地址
[email protected] SKILL.md:30 Info Email 邮箱地址
[email protected] SKILL.md:36 Info Email 邮箱地址
[email protected] SKILL.md:81 Info Email 邮箱地址
[email protected] SKILL.md:152 File Tree
1 files · 24.8 KB · 695 lines Markdown 1f · 695L
└─
SKILL.md
Markdown
Security Positives
✓ Pure documentation skill - no executable code present
✓ Provides legitimate security hardening patterns for AI agents
✓ No scripts, dependencies, or external calls
✓ IOCs flagged (rm -rf /, evil.com URLs) are defensive examples explaining threat scenarios, not actual malicious behavior
✓ Contains comprehensive security guidance including prompt injection defense, data boundaries, and WAL protocol
✓ All examples showing malicious patterns are clearly marked as what NOT to do