The Agent Skills Directory

[COMMAND_EXECUTION]: The toolkit includes several utility scripts (such as change-archive.mjs, context-verify.mjs, and lib/index.mjs) that execute local shell commands like git and zip via Node.js execSync. These operations are restricted to the local environment and are used for legitimate repository management tasks like archiving change workspaces or detecting documentation drift. Command arguments derived from user input, such as change titles, are sanitized using alphanumeric filters to prevent shell injection.
[SAFE]: The toolkit implements its own dependency-free YAML parser and serializer in lib/index.mjs to avoid risks associated with third-party supply chain vulnerabilities. The scripts enforce strict directory scoping to the .changes folder within the project root. Furthermore, the 'adversarial' stance defined in the challenge protocol is a documented behavioral instruction intended to improve software architecture quality and does not constitute a prompt injection or safety bypass attempt.

brainstorm