The Agent Skills Directory

[SAFE]: The skill operates entirely on user-provided fiction samples to create local style profiles. It includes explicit instructions in SKILL.md and references/novel_style_schema.md to avoid preserving long source passages or copyrighted content (character names, plot inventions), focusing instead on abstract prose mechanics.
[COMMAND_EXECUTION]: The skill uses Python scripts (novel_style_validator.py, novel_style_runtime_loader.py) for data processing. These scripts perform standard JSON parsing and file reading tasks within the skill's directory and do not execute arbitrary shell commands or access sensitive system paths.
[DATA_EXPOSURE]: The skill manages its own data under a novel_styles/ directory. It does not attempt to access sensitive environment variables, credentials, or network resources. The workflow specifically asks for user permission and source paths for local text files.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted user text samples for analysis, it uses structured prompts (prompts/style_analyzer.md, prompts/intake.md) to distill specific linguistic patterns. The risk of injection is low as the output is a structured JSON profile describing syntax and diction rather than executable instructions, and the skill includes 'Anti-AI' rules to filter out generic or synthetic patterns.

forge-novel-style