The Agent Skills Directory

[SAFE]: The skill is designed to bring projects into compliance with the 'Meta-O' contract, following strict guidelines for business framing, architecture decisions, and quality control. It enforces security best practices such as redacting secrets and using byte-identical agent instructions.\n- [COMMAND_EXECUTION]: The skill utilizes standard system utilities (git, jq, flock) and specialized ecosystem tools (orca, orca-cli). It operates under a least-privilege model, performing setup repairs in dedicated feature/meta-o-setup branches and requiring human approval for registrations or CI/CD modifications.\n- [DYNAMIC_EXECUTION]: The bundled scripts/mo-posture.sh script performs environment diagnostics by executing internal probe logic across multiple shell startup modes. The script is highly hardened; it uses privileged mode (-p) to ignore inherited functions, verifies that core builtins are not shadowed by aliases or malicious functions, and specifically refuses to expose sensitive data such as secrets or environment variable values.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes project configuration files (package.json, pyproject.toml) and CI/CD definitions. It mitigates potential injection from these sources by using structured parsers (js-yaml) instead of text-based patterns and by treating dynamic or unreadable remote CI constructs as 'unknown' rather than executing or trusting them.

mo-setup