The Agent Skills Directory

[SAFE]: The skill performs expected file operations, including reading raw material and writing or updating an article file, as part of its core writing assistance functionality. No unauthorized file access or exfiltration patterns were detected.
[PROMPT_INJECTION]: The instructions do not contain phrases or patterns aimed at overriding agent behavior, bypassing safety guidelines, or extracting system prompts. While the skill ingests untrusted raw material, which presents an indirect injection surface, the instructions are narrowly scoped to content generation and include a human-in-the-loop checkpoint where the user must approve beats before they are written.
[COMMAND_EXECUTION]: No shell commands, system calls, or subprocess executions are defined within the instructions or supporting documentation.

writing-beats