Skip to content

Security: Sunnylincc/ClawSale

Security

docs/security.md

Security and safety

Trust model

  • Gateway executes only registered skills.
  • Skill I/O is schema-validated with Zod.
  • Workspace data is local-first for explicit control.

Skill safety guidance

  • Skills should only use public/provided signals.
  • Do not attempt bypasses for private data.
  • Keep outbound generation human-reviewed.

Local development cautions

  • Workspace project files may contain sensitive sales notes.
  • Add stronger secret management before production deployment.

Data handling boundaries

ClawSale focuses on discovery, enrichment, scoring, and drafting. It intentionally avoids default spam/bulk-send automation.

There aren’t any published security advisories