The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to manage a local file system resource. It asks the user for a file path and performs periodic read and append operations to that Markdown file to store brainstorming notes.
[PROMPT_INJECTION]: The skill instructions create a surface for indirect prompt injection. By requiring the agent to 're-read [the file] from disk' before writing, the agent ingests data that could potentially contain instructions intended to influence its behavior. This is a common characteristic of document-editing tools and no specific exploits are identified.
[SAFE]: The skill's behavior is transparent and aligns with its stated purpose. There are no external dependencies, remote code downloads, or attempts to exfiltrate sensitive data.

writing-fragments