The Agent Skills Directory

[DATA_EXFILTRATION]: The skill hardcodes absolute local filesystem paths located in a user's home directory (/Users/jcohen/Obsidian/...). While these point to the author's own writing projects, hardcoding absolute paths to user directories is a poor security practice that exposes local folder structures and can lead to unintended data exposure if the environment changes.
[DATA_EXFILTRATION]: The skill is designed to accept arbitrary file paths from the user (argument-hint: "[text or file path to rewrite]"). Combined with the Read tool, this allows the agent to read any file on the local system that the user specifies.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted data (user-provided text or files) and incorporates it into the agent's reasoning process.
Ingestion points: User-provided text or files via arguments; local files read via the Read tool (SKILL.md).
Boundary markers: None. The skill does not use delimiters or instructions to isolate the writing style rules from the untrusted content being rewritten.
Capability inventory: The skill has access to the Read and Glob tools for filesystem operations.
Sanitization: No sanitization, validation, or filtering is performed on the input text or the content of the files read from the local system.

book-prose-writer