setup-linux-host-collection
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill implements a sophisticated defense against indirect prompt injection by introducing a 'wrap' utility. This helper uses unique nonces to delimit untrusted data from system logs and command outputs, ensuring that malicious content cannot be interpreted as instructions by the AI agent.
- [CREDENTIALS_UNSAFE]: The skill adheres to strict security protocols for secret management. It instructs users to utilize local shell environment variables for ingest tokens, ensuring that sensitive credentials are never exposed to the AI model's context or history.
- [EXTERNAL_DOWNLOADS]: The configuration instructions utilize the vendor's official repositories and well-known package hosting services for software delivery. These sources are consistent with the skill's stated purpose and the author's identity.
- [COMMAND_EXECUTION]: The skill contains administrative commands for package management and service configuration. It includes clear architectural boundaries and warnings that ensure these high-privilege operations are performed manually by the user on the target host rather than autonomously by the agent.
Audit Metadata