The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell commands for project maintenance, such as creating directories (mkdir -p), verifying file content (head), searching for registration entries (grep), and checking file existence (ls). These are all restricted to the internal .agents/skills/ directory structure and represent standard documentation workflows.
[PROMPT_INJECTION]: While the skill discusses how agents use the description field for routing, it does not include instructions to override safety guidelines or bypass constraints. The guidance focuses on clarity and functional discoverability.
[DATA_EXPOSURE]: The skill interacts exclusively with repository metadata and documentation files (e.g., SKILL.md, README.md). It does not access sensitive user directories, credentials, or environment variables.
[REMOTE_CODE_EXECUTION]: There are no remote downloads, package installations, or external script executions. All operations are local to the repository environment.

writing-skills