The Agent Skills Directory

[COMMAND_EXECUTION]: Step 0 in SKILL.md contains a bash script that searches the user's $HOME/.claude/plugins/cache/ directory for an executable named mthds-env-check and runs it using exec. This constitutes dynamic loading and execution from a computed file path.
[COMMAND_EXECUTION]: The skill uses Python heredocs to generate and execute scripts at runtime for synthetic document generation (PDF, Word, and Excel files). This dynamic script generation and execution occurs in several sections: 'PDF Documents', 'Word Documents (DOCX)', and 'Spreadsheets (XLSX)'.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the mthds package globally using npm install -g mthds if the environment check fails.
[PROMPT_INJECTION]: The skill implements an 'Indirect Prompt Injection' surface by ingesting and processing untrusted user data (PDFs, images, and documents).
Ingestion points: Files provided by the user via paths or folders in Step A and Step B.
Boundary markers: None mentioned for the interpolation of user file content into prompts.
Capability inventory: Uses Bash, Read, Write, and mthds-agent for processing.
Sanitization: No explicit sanitization or validation of the content within the user-provided files is described.

mthds-inputs