The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a shell script scripts/check_site_hygiene.sh intended to scan the local repository for common secret signatures and sensitive file paths (like .env or private keys) before sensitive actions such as commits or deployments.\n- [OBFUSCATION]: The hygiene script uses minor string concatenation (e.g., 'gh''p_') for its internal signature patterns. This is a common defensive practice to prevent the security script itself from being flagged by other scanners during distribution or versioning.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external Markdown content and site configuration files, which are ingestion points for potentially untrusted data.\n
Ingestion points: Processes files in the repository including docs/, zensical.toml, and pyproject.toml.\n
Boundary markers: The skill provides explicit instructions to the agent to separate editorial intent from implementation and to treat any hygiene findings as signals for manual review.\n
Capability inventory: Shell access is used for localized build and validation tasks; file system access is used for requested edits.\n
Sanitization: The skill mandates human-authorized changes and the use of the project's own validation tools, reducing the risk of autonomous execution of embedded instructions.

zensical