AST10: Cross-Platform Reuse
Risk Level: Medium
A skill written for one agent platform gets ported to another, and the security properties do not survive the trip. A permission manifest that the source platform enforced may have no equivalent field in the destination format, so it silently disappears. The skill keeps working. The control does not.
Why it persists
Snyk confirmed malicious skills published simultaneously to ClawHub and skills.sh by the same threat actors, with neither platform sharing scanning intelligence with the other. Without a universal format and shared threat intelligence, each registry re-learns the same campaign independently, and multi-platform stacks have no unified governance.
Mitigations
- Re-validate full security metadata whenever a skill is ported to another platform
- Normalize risk tier, permission, and signature fields across the platforms you support
- Build scanners that evaluate skill content independently of any one runtime
- Adopt a universal skill format for new development instead of per-platform variants
- Participate in cross-registry threat intelligence sharing rather than trusting per-platform scans