The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes platform discovery commands (uname, sysctl, arch) and the local CLI tool (repo-native-alignment) to identify the environment and configure the project.
[SAFE]: Implements secure configuration practices by explicitly instructing the agent to use only repo-local, checked-in, and vendored assets for setting up hooks and agents, rather than fetching mutable remote content.
[PROMPT_INJECTION]: The skill is subject to potential indirect prompt injection as it reads arbitrary project files (like README or package manifests) to generate summaries. This risk is effectively mitigated by a design that requires explicit user review and approval of the generated summary before it is written to the persistent AGENTS.md file.
[SAFE]: Project metadata and file structures are scanned to synthesize project context. This activity is restricted to local file access, and the resulting context is stored within the project's own repository files.

teach-oh