skills/mukul975/anthropic-cybersecurity-skills/exploiting-active-directory-with-bloodhound/Gen Agent Trust Hub
exploiting-active-directory-with-bloodhound
Warn
Audited by Gen Agent Trust Hub on Mar 15, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The script
scripts/agent.pyincludes a hardcoded default password (bloodhound) for the Neo4j database connection used to store and query Active Directory data. - [COMMAND_EXECUTION]: The file
scripts/agent.pyinvokes external security collectors (SharpHound.exeandbloodhound-python) viasubprocess.check_output. While these are legitimate tools for the skill's stated purpose, the execution of external binaries poses a risk in automated environments. - [EXTERNAL_DOWNLOADS]: The
references/workflows.mdfile contains a command to download a configuration file from an external URL (https://ghst.ly/getbhce), which is an unverified source for security-sensitive materials. - [DATA_EXFILTRATION]: The core functionality of the skill involves the collection and extraction of sensitive Active Directory relationship data, including group memberships, user properties, and session information. This behavior is the primary objective of the skill and represents a significant data handling risk.
- [PROMPT_INJECTION]: The skill ingests untrusted Active Directory data from JSON and ZIP files (in
scripts/agent.pyandscripts/process.py) to generate reports and findings. The lack of data sanitization or boundary markers represents a surface for indirect prompt injection, where malicious AD objects could influence the agent's summary and output.
Audit Metadata