審計紀錄 - coordinator

審計版本 6

高風險問題 (4)

SKILL.md:10-18 SKILL.md:21-43 SKILL.md:117-155

Autonomous Offensive Testing Workflow

The skill directs agents to perform reconnaissance, exploit services, escalate privileges, and capture user and root flags. This is dual-use security content, but the activation on a target IP and the exploitation workflow create high risk outside a controlled lab.

AD_ATTACK_SUPPLEMENT.md:3-6 AD_ATTACK_SUPPLEMENT.md:24-33 AD_ATTACK_SUPPLEMENT.md:528-540 SKILL.md:78-90

Prompt Injection Attempt Detected

The files include autonomy override language such as "ABSOLUTE RULE: You CANNOT stop execution" and "NEVER ask for help". These instructions try to control the agent loop and could conflict with user control, safety review, or operational limits.

AD_ATTACK_SUPPLEMENT.md:77-105 AD_ATTACK_SUPPLEMENT.md:130-170 AD_ATTACK_SUPPLEMENT.md:249-309 AD_ATTACK_SUPPLEMENT.md:324-388

Active Directory Attack Playbook

The supplement provides concrete commands for empty-password authentication, LDAP modification, Kerberos ticket attempts, SMB execution, account creation, ACL modification, and SPN abuse. These are offensive techniques that can affect real networks if used without authorization.

AD_ATTACK_SUPPLEMENT.md:207-214 AD_ATTACK_SUPPLEMENT.md:468-492

Credential Attack and Weak Hash Workflow

The supplement instructs conversion of a found password to an NTLM hash using MD4 and then using that hash for authentication. MD4 and NTLM are weak in defensive cryptography, and here they are used to support credential abuse.

中風險問題 (2)

SKILL.md:21-45 SKILL.md:95-98 AD_ATTACK_SUPPLEMENT.md:77-106 AD_ATTACK_SUPPLEMENT.md:249-309

External Command Instructions in Documentation

The static Ruby backtick findings are false positives because the backticks are Markdown fences. However, the fenced blocks contain commands for netexec, impacket, ldapsearch, ldapmodify, xfreerdp, and similar tools, so the operational guidance remains risky.

SKILL.md:49-52 SKILL.md:94-98

Local State File Dependency

The skill instructs the agent to read a local .pentest-state.json file to guide decisions. This is a limited filesystem interaction, but it could expose captured flags or target state if run in a shared workspace.

低風險問題 (2)

SKILL.md:25 SKILL.md:120-125

Placeholder Target IP Addresses

The hardcoded IP detections refer to 10.10.10.1 in examples. These appear to be placeholder lab targets rather than fixed exfiltration endpoints, although the examples still normalize network scanning against supplied hosts.

SKILL.md:1-4 AD_ATTACK_SUPPLEMENT.md:544-563

No Packaged Executable Malware Found

No evidence found of bundled scripts that automatically execute malware, persist on the host, or exfiltrate local secrets. The risk comes from agent instructions and offensive command guidance rather than installed binary payloads.

風險因素

⚙️ 外部命令 (50)

🌐 網路存取 (3)

SKILL.md:25 SKILL.md:120 SKILL.md:125

📁 檔案系統存取 (3)

SKILL.md:49-52 SKILL.md:94-98 SKILL.md:123

偵測到的模式

Delegated Exploitation PromptsNetwork Reconnaissance and Authentication CommandsDirectory Modification and Account Abuse CommandsInfinite Retry Loop Guidance

審計版本 5

中風險

Jan 16, 2026, 09:47 PM

This skill is a strategic coordinator for authorized penetration testing in controlled environments (CTF, training). Static findings show documentation of attack techniques in AD_ATTACK_SUPPLEMENT.md and SKILL.md, but these are educational examples, not executable code. The skill delegates all execution to subagents and explicitly restricts usage to authorized environments like HackTheBox and VulnHub.

3

已掃描檔案

958

分析行數

2

發現

claude

審計單位

中風險問題 (1)

AD_ATTACK_SUPPLEMENT.md:1-564 SKILL.md:1-196

Attack Technique Documentation

The skill contains educational documentation of penetration testing techniques including Active Directory exploitation, password attacks, and privilege escalation methods. Intended for authorized environments (CTF, training).

風險因素

⚡ 包含腳本 (3)

AD_ATTACK_SUPPLEMENT.md:77-106 AD_ATTACK_SUPPLEMENT.md:130-171 AD_ATTACK_SUPPLEMENT.md:191-237

審計版本 4

中風險

Jan 16, 2026, 09:47 PM

This skill is a strategic coordinator for authorized penetration testing in controlled environments (CTF, training). Static findings show documentation of attack techniques in AD_ATTACK_SUPPLEMENT.md and SKILL.md, but these are educational examples, not executable code. The skill delegates all execution to subagents and explicitly restricts usage to authorized environments like HackTheBox and VulnHub.

3

已掃描檔案

958

分析行數

2

發現

claude

審計單位

中風險問題 (1)

AD_ATTACK_SUPPLEMENT.md:1-564 SKILL.md:1-196

Attack Technique Documentation

The skill contains educational documentation of penetration testing techniques including Active Directory exploitation, password attacks, and privilege escalation methods. Intended for authorized environments (CTF, training).

風險因素

⚡ 包含腳本 (3)

AD_ATTACK_SUPPLEMENT.md:77-106 AD_ATTACK_SUPPLEMENT.md:130-171 AD_ATTACK_SUPPLEMENT.md:191-237

審計版本 3

中風險

Jan 10, 2026, 12:06 PM

This skill contains penetration testing methodologies for educational use in controlled environments. While it documents attack techniques, it's designed for legitimate security research and training purposes. The skill delegates actual execution to other agents and maintains ethical boundaries.

2

已掃描檔案

760

分析行數

1

發現

claude

審計單位

中風險問題 (1)

AD_ATTACK_SUPPLEMENT.md:64-563

Penetration Testing Attack Documentation

The skill contains detailed documentation of various attack techniques including Active Directory exploitation, password attacks, and privilege escalation methods. While intended for educational purposes in controlled environments, this information could be misused if accessed by malicious actors.

審計版本 2

中風險

Jan 10, 2026, 12:06 PM

This skill contains penetration testing methodologies for educational use in controlled environments. While it documents attack techniques, it's designed for legitimate security research and training purposes. The skill delegates actual execution to other agents and maintains ethical boundaries.

2

已掃描檔案

760

分析行數

1

發現

claude

審計單位

中風險問題 (1)

AD_ATTACK_SUPPLEMENT.md:64-563

Penetration Testing Attack Documentation

The skill contains detailed documentation of various attack techniques including Active Directory exploitation, password attacks, and privilege escalation methods. While intended for educational purposes in controlled environments, this information could be misused if accessed by malicious actors.

審計版本 1

中風險

Jan 10, 2026, 12:06 PM

This skill contains penetration testing methodologies for educational use in controlled environments. While it documents attack techniques, it's designed for legitimate security research and training purposes. The skill delegates actual execution to other agents and maintains ethical boundaries.

2

已掃描檔案

760

分析行數

1

發現

claude

審計單位

中風險問題 (1)

AD_ATTACK_SUPPLEMENT.md:64-563

Penetration Testing Attack Documentation

The skill contains detailed documentation of various attack techniques including Active Directory exploitation, password attacks, and privilege escalation methods. While intended for educational purposes in controlled environments, this information could be misused if accessed by malicious actors.