skills/mukul975/anthropic-cybersecurity-skills/eradicating-malware-from-infected-systems/Gen Agent Trust Hub
eradicating-malware-from-infected-systems
Warn
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The Python scripts
scripts/agent.pyandscripts/process.pyutilize thesubprocessmodule to execute high-privilege commands includingtaskkill,reg delete,schtasks, andsc stop/deleteto eradicate malware artifacts. - [EXTERNAL_DOWNLOADS]: The skill integrates with the CrowdStrike Falcon API (
api.crowdstrike.com) for post-eradication scanning and references standard tools such as YARA, ClamAV, and Sysinternals Autoruns. - [DATA_EXFILTRATION]: The skill accesses sensitive system files and configuration data, including
~/.ssh/authorized_keys, crontabs, and Windows Registry run keys, to identify unauthorized entries. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted data from the filesystem and registry. Evidence chain: (1) Ingestion points:
scripts/process.pyreads registry keys and shell profiles; (2) Boundary markers: None present; (3) Capability inventory: Process termination and file deletion viasubprocess; (4) Sanitization: None identified for external content interpolation into reports.
Audit Metadata