skills/galyarderlabs/galyarder-framework/eradicating-malware-from-infected-systems/Gen Agent Trust Hub
eradicating-malware-from-infected-systems
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill documentation includes a 'GLOBAL PROTOCOLS' section that uses directive language ('MANDATORY', 'Bypass planning') to define the agent's cognitive processes and operational modes. This includes instructions to bypass planning for hotfixes and mandates specific logic loops, which attempts to override default agent behavior.- [COMMAND_EXECUTION]: The provided scripts ('agent.py' and 'process.py') and the workflow in 'SKILL.md' utilize system-level commands to manage the host environment. This includes process termination using 'taskkill' and 'kill', registry modification via 'reg', and service management using 'systemctl' and 'sc'.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the host system, such as registry keys, scheduled task names, and file contents, to identify artifacts for removal. In the context of an infected system, this data is attacker-controlled and could influence agent logic due to the lack of boundary markers or rigorous sanitization during parsing.
- Ingestion points: Registry query results, scheduled task CSV outputs, and startup file contents parsed in 'scripts/agent.py' and 'scripts/process.py'.
- Boundary markers: None; system data is interpolated directly into reporting and removal logic.
- Capability inventory: The skill possesses extensive capabilities for file system deletion, process management, and registry modification.
- Sanitization: Values are truncated for display in 'process.py', but the original data is used for identification and execution actions.- [PRIVILEGE_ESCALATION]: The instructions describe incident response actions that require administrative or root privileges, such as resetting the Domain Controller's Kerberos Ticket Granting Ticket (KRBTGT) password and performing directory-wide credential rotations.- [EXTERNAL_DOWNLOADS]: The workflow incorporates network operations to a well-known security service, making API calls to CrowdStrike's scanning endpoint to validate eradication success. These operations are documented as part of the intended security workflow.
Audit Metadata