The Agent Skills Directory

[OBFUSCATION]: The skill employs string concatenation in SKILL.md to hide sensitive identifiers and evade signature-based detection by security software.\n
Evidence: The PowerShell script splits strings like 'Am' + 'siUtils' and 'am' + 'siInitFailed' to avoid triggering security alerts during bypass operations.\n- [PRIVILEGE_ESCALATION]: The skill provides explicit instructions and code to bypass the Antimalware Scan Interface (AMSI) on Windows, which is a key security control.\n
Evidence: The script uses .NET reflection to disable AMSI initialization failures, allowing unauthorized scripts to run.\n- [EXTERNAL_DOWNLOADS]: The skill recommends downloading and running a remote setup file via a URL shortener which can lead to execution of unverified content.\n
Evidence: `curl -L https://ghst.ly/getbhce | docker compose -f
up` in SKILL.md downloads and immediately executes a Docker configuration.\n- [COMMAND_EXECUTION]: The skill relies on the execution of several security-sensitive command-line tools and service management commands.\n
Evidence: Use of SharpHound.exe, azurehound, and sudo neo4j for Active Directory enumeration and database management.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data generated by external Active Directory enumeration tools, creating a vulnerability surface for malicious input.\n
Ingestion points: scripts/process.py loads user, computer, and group data from JSON files; scripts/agent.py retrieves data from Neo4j graph queries.\n
Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands in the processed data.\n
Capability inventory: The scripts have the ability to write files (json.dump in scripts/agent.py), list and delete directories (shutil.rmtree in scripts/process.py), and output data to the console.\n
Sanitization: Absent; the ingested data is processed directly without verification or filtering.

performing-active-directory-bloodhound-analysis