The Agent Skills Directory

[SAFE]: The skill is focused on text processing and content formatting. It provides instructions for structural adjustments (splitting paragraphs, creating lists) and stylistic constraints (rounding numbers, avoiding specific phrases).
[COMMAND_EXECUTION]: The skill suggests using grep to identify specific voice violations in local files. This is a benign use of standard file-searching tools for editing purposes.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data (article drafts), which represents a potential attack surface. However, the skill restricts the agent's actions to text transformation and reporting changes to a human reviewer ('Akita') before proceeding, significantly mitigating the risk. No high-privilege capabilities such as network access or arbitrary code execution are requested or used.
[METADATA_POISONING]: The metadata and instructions are consistent with the stated purpose of finalizing blog content.

akita-finish