The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/agent.py and scripts/process.py scripts utilize the subprocess module to execute system commands for backup auditing. agent.py runs restic to inspect snapshots and repository integrity, while scripts/process.py uses the aws CLI to query S3 Object Lock configurations. These executions are functional and scoped to specific, pre-defined auditing tasks.
[EXTERNAL_DOWNLOADS]: The skill documentation and reference files mention external dependencies and services. The references/api-reference.md file recommends installing the boto3 library, and scripts/agent.py imports the requests library. The skill also references official documentation and endpoints for AWS, Azure, and Veeam, which are well-known and trusted providers.
[INDIRECT_PROMPT_INJECTION]: The audit scripts exhibit a surface area for indirect prompt injection as they ingest data from external sources like the Veeam API and Restic/AWS CLI outputs.
Ingestion points: Data is fetched from the Veeam REST API and the output of Restic and AWS CLI commands.
Boundary markers: The scripts do not use specific delimiters or warnings to isolate external data from the audit logic.
Capability inventory: The scripts can execute subprocesses, make network requests, and write report files.
Sanitization: Minimal sanitization is performed (e.g., JSON parsing and string slicing). However, the scripts do not perform dangerous actions based on the ingested content, maintaining a low risk level.
[SAFE]: The skill adheres to security best practices for ransomware defense, including credential isolation, the use of MFA, and the implementation of immutable storage. Secret management in code snippets correctly utilizes environment variables, and administrative commands like sudo chattr are used appropriately for system hardening.

implementing-ransomware-backup-strategy