The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like ls, stat, mkdir, mv, and rm -rf to manipulate files. * Evidence: SKILL.md Steps 2 and 3. * Risk: Filenames are interpolated into commands without explicit escaping instructions, potentially leading to command injection if filenames contain shell metacharacters.
[PROMPT_INJECTION]: The skill processes filenames from the local filesystem to determine its logic, creating a surface for indirect prompt injection. * Ingestion points: SKILL.md Step 2.1 uses ls -R to read untrusted filenames. * Boundary markers: No delimiters or ignore instructions are used for filenames. * Capability inventory: Access to filesystem commands mv and rm -rf. * Sanitization: No instruction to sanitize or validate filenames before they are used in command generation.

media-organizer