The Agent Skills Directory

[SAFE]: The skill's primary functionality is limited to reading an input file provided by the user and writing the resulting article to a new file. These are standard operations for a document-shaping tool.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted raw material (the 'input pile'). While an attacker could potentially embed malicious instructions within this source material to influence the generated article, the skill's instructions require the agent to 'rework' and 'mine' the fragments into a specific structure, which provides a natural layer of filtering. Furthermore, the skill lacks high-risk capabilities such as network access or shell command execution, ensuring that even a successful prompt injection would have minimal security impact.
[SAFE]: All metadata and instructions are consistent with the stated purpose of assisting in long-form writing and structure commitment.

writing-shape