secure-skill-runtime

Prevent skill overwrite attacks by loading a runtime security guard that enforces explicit human approval for external changes before any content persists to the skill store, ensuring only validated commits modify skills.

Provenance tracking for agent skills records the origin and approval history of external content. You need it whenever an agent processes external content or writes to the skill store to ensure only approved sources persist.

Block untrusted repositories using automated no-go repo management, which maintains a blocked source list and validates content origins during runtime security audits to prevent untrusted sources from persisting.

Runtime security for agent skills detects denial of service patterns by enforcing resource limits and validation during external content processing, preventing state corruption and exploitation through automated checks.

You can audit runtime security and check for state corruption by loading the security guard during provenance checks, which validates that only approved content persists and flags any unauthorized modifications.

The runtime security guard requires loading as part of a secure-* skill sequence and depends on explicit human approval for external changes, meaning automated workflows may pause for validation during skill store writes.