The Agent Skills Directory

[SAFE]: The skill implements a security-focused 'Fabric Schema' loop designed to prevent unintended file modifications. It uses a hypothesize-verify-commit sequence with short-lived certificates and workspace fingerprinting to ensure integrity.
[COMMAND_EXECUTION]: The skill mentions the use of 'trusted commands' and provided examples using pnpm exec vitest. These commands are described as being part of static host configuration, which limits the risk of arbitrary command injection. These executions are intended for behavior-preservation checks and verification within a local development environment.

fabric-schema