gog-safety

What This Skill Does

The gog-safety skill provides a secure framework for building and deploying gogcli binaries with compile-time command removal. By stripping out unauthorized functionality at the binary level rather than the runtime level, it ensures that forbidden operations (such as sending unauthorized emails or performing administrative tasks) are physically impossible for the AI agent to execute. This is achieved by utilizing PR #366 from the steipete/gogcli repository, which leverages a YAML-based configuration to whitelist specific subcommands before compilation.

Installation

To install this skill, use the ClawHub CLI command: clawhub install openclaw/skills/skills/brennerspear/gog-safety. Ensure you have Go 1.22+ and git installed on your build machine, as the process involves cloning the source, applying security profiles, and compiling the binary from scratch.

Use Cases

This skill is ideal for security-conscious organizations deploying AI agents that require access to internal messaging and document tools. Choose L1 for agents performing automated email triage and drafting, L2 for agents managing RSVP and collaborative feedback loops, or L3 for agents that require full writing capabilities while explicitly blocking dangerous administrative overrides.

Example Prompts

1. "Build a version L1 binary for my local ARM64 machine and prepare it for deployment to the staging server."
2. "Deploy the new L2 safety profile to the target host and run a verification check to ensure that the gmail send command is successfully disabled."
3. "Revert the current host binary to the previous state using the gog-backup file after the latest L3 update caused a workflow conflict."

Tips & Limitations

Always use the --verify flag during deployment to ensure your security policy is enforced. Note that while compile-time removal is the gold standard for security, certain architectural edge cases remain: for example, filter forwarding and drive sharing are permitted at the L1 level to allow for necessary inbox organization. These are accepted design risks for version 1.0. Always keep track of your security profiles, as custom YAML modifications allow you to tailor the agent's capabilities precisely to your business requirements.