joelclaw-web

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions include running system commands like kubectl, tailscale, and launchctl to gather status for a public monitoring page. These are necessary for the skill's stated purpose and are managed under specific security rules.
  • [PROMPT_INJECTION]: The skill identifies an ingestion surface for indirect injection via processed content.
  • Ingestion points: Data from Convex database and local MDX files.
  • Boundary markers: No explicit delimiters for data interpolation are specified.
  • Capability inventory: Access to the local monorepo filesystem, database upserts, and an internal revalidation API.
  • Sanitization: Uses regex for specific structural formatting, but lacks general sanitization for embedded instructions.
  • [DATA_EXFILTRATION]: To prevent exposure of infrastructure secrets, the skill mandates 'OPSEC Rules' that require aliasing hostnames and stripping sensitive identifiers (ports, IPs) before data is rendered publicly.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 10:25 AM