ad-cs-abuse

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads tools from several external GitHub repositories that are not recognized as trusted sources, including topotam/PetitPotam, p0dalirius/Coercer, bats3c/ADCSPwn, trimarcjake/Locksmith, and zynnnnnn/ADCSToolkit. Additionally, a URL flagged by automated scanners (http://attacker.example.com/cert.crt) is included as an example in payloads.md for certificate chain manipulation.
  • [REMOTE_CODE_EXECUTION]: The instructions in SKILL.md and payloads.md direct the agent to clone repositories from third-party sources and execute pip install -r requirements.txt to install their dependencies, which constitutes a remote code execution risk from unverified sources.
  • [DATA_EXFILTRATION]: The skill contains multiple payloads and methodology steps for extracting sensitive domain credentials and NTDS hashes using impacket-secretsdump, which is a common data exfiltration pattern in credential theft attacks.
  • [COMMAND_EXECUTION]: The skill relies heavily on the execution of complex shell commands for Active Directory reconnaissance and exploitation, involving tools like certipy, rpcclient, and ntlmrelayx to modify domain infrastructure and impersonate privileged accounts.
  • [PERSISTENCE]: Detailed instructions are provided for establishing persistence using "Shadow Credentials" by writing to the msDS-KeyCredentialLink Active Directory attribute (Event ID 4662), allowing an attacker to maintain access that survives password resets.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by processing external data from Active Directory without sufficient safeguards.
  • Ingestion points: The skill ingests untrusted data from LDAP search results (templates, enrollment services), CA database records, and certificate metadata.
  • Boundary markers: There are no explicit delimiters or instructions provided to ensure the agent ignores potentially malicious content embedded within the processed AD data.
  • Capability inventory: The skill utilizes significant capabilities, including shell command execution via the Bash tool, Active Directory object modification, and network operations.
  • Sanitization: The methodology does not include steps to sanitize or validate the external data retrieved from the Active Directory environment before processing it.
Recommendations
  • Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM