The Agent Skills Directory

[SAFE]: The skill implements standard file-handling logic to manage local markdown documents. It does not perform any unauthorized network requests or access sensitive system paths by default.
[PROMPT_INJECTION]: There is a minimal surface for indirect prompt injection because the skill is instructed to re-read the fragment file from disk before each write. This behavior is a functional requirement to preserve user edits but technically allows existing content to influence the agent. Ingestion points: User-provided text and existing content from the specified markdown file. Boundary markers: Individual fragments are separated by horizontal rules (---). Capability inventory: The skill possesses file read and write permissions within the user's environment. Sanitization: The skill appends raw user input without specific escaping or filtering.

writing-fragments