The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py uses subprocess.run to execute system commands such as systemctl on Linux and sc on Windows to verify that restored services are active. These commands are executed without a shell (shell=False) and are a core functional requirement for the recovery validation process.
[INDIRECT_PROMPT_INJECTION]: The skill processes external JSON configuration files and hash manifests which could potentially be manipulated by an attacker to influence agent behavior. 1. Ingestion points: The run_recovery_drill function in scripts/agent.py loads system definitions, service names, and file paths from a user-provided JSON config file. 2. Boundary markers: None identified; the skill assumes configuration files are provided by a trusted administrator within an isolated recovery environment. 3. Capability inventory: The script performs file system reads (recursive hashing), network connectivity checks (TCP sockets), and command execution (service status queries). 4. Sanitization: The script performs standard path existence checks but does not implement complex sanitization for service names or hostnames.
[SAFE]: All identified capabilities—including system status checks, directory hashing, and network port validation—align perfectly with the stated purpose of testing ransomware recovery resilience. The instructions explicitly recommend using an isolated test environment, adhering to security best practices.

testing-ransomware-recovery-procedures