The Agent Skills Directory

[SAFE]: The skill implements a structured workflow for shipping code through GitHub PRs and deployments. It focuses on using established tools (git, gh) and project-specific commands rather than introducing new external dependencies.
[SAFE]: Data privacy is a core component, with explicit instructions under 'Publication privacy' to sanitize PR descriptions and exclude sensitive identifiers, personal data, and internal URLs before publication.
[SAFE]: The skill uses local verification steps (e.g., hard-eng.py ship --stage ready) to ensure that PRs meet requirements before merging. It also mandates runtime verification for deployments using project-defined commands.
[SAFE]: Cleanup procedures are designed to be non-destructive, specifically preserving untracked files, submodules, and worktrees that it does not explicitly own. It uses python3 -B to avoid generating new bytecode artifacts during the process.
[SAFE]: It avoids dangerous patterns such as arbitrary remote code execution or history rewriting, requiring separate authorization for high-risk Git operations.

he-ship