The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it ingests and summarizes untrusted data from user-provided URLs and raw files.
Ingestion points: Content fetched from user-supplied URLs and files located in the ~/Personal_wiki/raw/ directory.
Boundary markers: Absent. There are no explicit instructions or delimiters defined to separate user data from system instructions during processing.
Capability inventory: The skill can create directories and files within the user's home directory and perform network requests to fetch external content.
Sanitization: Absent. The skill does not describe any validation or sanitization of external text before it is integrated into the wiki.
[REMOTE_CODE_EXECUTION]: The skill provides a Python code template for OCR tasks that uses pytesseract and pdf2image. This involves the runtime execution of script content to process local files.
[COMMAND_EXECUTION]: The skill uses standard shell utilities, including mkdir, ls, grep, and tail, to initialize and manage the wiki's file structure and logging system within the ~/Personal_wiki/ directory.
[EXTERNAL_DOWNLOADS]: The skill references a PDF processing tool located in a public repository belonging to a well-known organization to handle document ingestion.

personal-wiki