The Agent Skills Directory

[COMMAND_EXECUTION]: The skill mandates the execution of various Python scripts residing in the user's home directory, specifically within a hidden folder structure.
Evidence: Found in the 'Verification Scripts (MANDATORY)' section of SKILL.md, which lists commands like python ~/.claude/skills/frontend-design/scripts/ux_audit.py . and python ~/.claude/skills/vulnerability-scanner/scripts/security_scan.py ..
Context: The skill instructs the agent to run these scripts after completing work, which constitutes execution of code from external local paths.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and modify project source code, creating a surface for indirect prompt injection.
Ingestion points: Project files processed by the agent, such as UserService.ts and other source files edited during tasks.
Boundary markers: Absent. The instructions do not define delimiters or provide 'ignore embedded instructions' prompts for the content being edited.
Capability inventory: The skill has access to Read, Write, and Edit tools, and it facilitates shell command execution through Python scripts.
Sanitization: Absent. There are no instructions for the agent to validate, escape, or sanitize the content of the files it is editing before performing operations based on them.

clean-code