The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the dynamic context injection syntax (!command) to run discovery commands like pwd and find at load time. These are used safely to determine the presence of project manifest files (e.g., package.json, Cargo.toml) and populate the agent's context with environment details.
[COMMAND_EXECUTION]: The skill uses scoped tools such as jq to parse local configuration files and the claude plugin CLI to interface with the plugin marketplace. These operations are restricted to relevant configuration data.
[SAFE]: The skill accesses sensitive local files like ~/.claude/settings.json to check for globally enabled plugins. This is consistent with its stated purpose of preventing redundant project-level configurations and does not involve any network exfiltration of this data.
[SAFE]: Remediation and file modification steps include safety measures, specifically the creation of a .backup file and mandatory user confirmation checkpoints before any changes are written to the filesystem.

health-audit