template-skill

This Skill automates generation of precise problem statements for vector-based retrieval by producing three-sentence prompts with structured JSON metadata. It enforces strict English, no filler, verb-first structure, and outputs valid JSON containing request, description, is_toxic, and toxic_reason fields for semantic search indexing and RAG pipelines.

The Skill outputs a valid JSON object with four required fields: request, description, is_toxic, and toxic_reason. Descriptions are exactly three declarative sentences starting with a verb or subject, with no filler language, designed for consistent indexing in RAG systems and agent decision workflows.

Yes, this template provides metadata-first scaffolding with mandatory YAML frontmatter for name and description. It includes structured guidance on resources and QA steps, enabling rapid prototyping while maintaining quality standards for new Skill development.

Problem statements must be exactly three sentences, strictly in English with no filler phrases. Each sentence must start with a verb or subject; the output must be valid JSON. These constraints ensure consistency across semantic search, RAG indexing, and agent workflows.

The Skill generates standardized, toxicity-flagged problem statements that feed directly into semantic search indexing and agent decision workflows. Structured JSON output with toxicity metadata enables reliable retrieval and filtering in production RAG systems.