The Agent Skills Directory

[SAFE]: The skill implements a robust, modular architecture for creative writing orchestration. It uses a clear separation of concerns by delegating specific writing tasks to specialized subagents through structured prompts.- [COMMAND_EXECUTION]: The skill legitimately executes local Python scripts (check_wordcount.py and check_chapter_quality.py) to perform automated quality control, such as monitoring word counts and detecting common AI writing fingerprints. These scripts use only standard libraries and do not perform network operations.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided inputs and reference materials. This ingestion surface is well-managed through the use of discrete task prompts and subagent role definitions. (1) Ingestion points: User queries and reference files stored in /mnt/agents/output/research/. (2) Boundary markers: Extensive use of create_subagent and task prompts to define strict operational boundaries. (3) Capability inventory: File system writes and local script execution. (4) Sanitization: Content is processed via LLM subagents within a structured workflow.

general-writing