The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external files in the memo_origin folder.
Ingestion points: Files within the memo_origin folder, including text, markdown, Word documents, and images.
Boundary markers: The instructions do not define delimiters or specific "ignore" markers for the content being processed.
Capability inventory: The skill utilizes shell command execution via the stat command and performs file-write operations to the memo/ directory.
Sanitization: No specific sanitization or validation of the input file content is mentioned.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (stat) using a placeholder for the file path ("<원본 파일 경로>"). If a file in the memo_origin directory has a maliciously crafted name containing shell meta-characters (e.g., ;, backticks, or $()), it could lead to command injection when the agent attempts to retrieve the file's metadata.

memo-sync