The Agent Skills Directory

[SAFE]: The skill is designed with security in mind. It uses helper scripts (in the scripts/ directory) to manage its runtime data, and these scripts explicitly validate all input to prevent path traversal and ensure all file operations are confined to the .ustht/ directory.
[PROMPT_INJECTION]: The skill processes untrusted user input and project-local markdown files, which represents an indirect prompt injection surface.
Ingestion points: User messages are captured by write_raw.py, and existing project files can be scanned via the /ustht import command.
Boundary markers: Captured thoughts are stored in structured date-based markdown headings with specific entry formats.
Capability inventory: The skill uses read, write, and bash tools to interact with local files through its bundled Python scripts.
Sanitization: The common.py module implements ensure_runtime_path and validate_dim_name to ensure that even if malicious content is processed, it cannot escape the designated storage directory or execute unauthorized system commands.

user-thoughts