satellite-rig

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill pipes local shell scripts (e.g., scripts/setup-satellite-rig.sh) directly into remote bash sessions over SSH using the cat | ssh 'bash -s' pattern. This facilitates remote configuration and management of target hosts.
  • [COMMAND_EXECUTION]: The skill makes extensive use of system commands and requires sudo privileges to perform administrative tasks, such as mounting NFS shares and installing system-wide persistence mechanisms via LaunchDaemons (/Library/LaunchDaemons/com.joelclaw.satellite.nas-mounts.plist).
  • [CREDENTIALS_UNSAFE]: Sensitive credentials like TYPESENSE_API_KEY and PDF_BRAIN_API_TOKEN are managed using a secrets lease utility. These keys are written to a local configuration file (~/.config/system-bus.env) with restricted filesystem permissions (0600).
  • [DATA_EXFILTRATION]: The skill interacts with internal network infrastructure, performing health checks and data searches against specific Tailscale nodes (tail7af24.ts.net) and local network assets (192.168.1.163).
  • [PROMPT_INJECTION]: The skill processes data from external sources, such as session search results and document summaries. This ingestion point serves as a potential surface for indirect prompt injection, although the provided verification examples use specific parsing to limit direct execution of untrusted content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 05:05 PM
Security Audit — agent-trust-hub — satellite-rig