es:skill-forge

To create hallucination-proof knowledge-graph skills, use a research-first workflow that mandates right vs wrong examples, authoritative references, and explicit common pitfalls in every reference file. This structured approach ensures auditable, decision-table-driven outputs.

The best way to structure a domain-specific skill is to design a standardized contract file alongside cohesive reference files. This workflow enforces decision tables, failure modes with exact symptoms, and strict 300-line file limits to ensure precise diagnosis.

A research methodology workflow prevents hallucination by requiring a structured audit of existing coverage, researching community solutions, and enforcing mirror and duplication validation across required directories before generating a consolidated knowledge graph.

Yes, you need YAML frontmatter containing the name and description fields to build a knowledge-graph skill. This metadata establishes the foundational skill contract required for the research, analyze, design, build, and validate workflow.

Skill validation fails when reference files lack mandatory right and wrong examples, decision tables, or explicit failure modes. It also fails if files exceed the 300-line limit or if mirror and duplication checks across required directories detect structural inconsistencies.

Yes, you can expand an existing knowledge-graph skill by applying the structured audit and research workflow to detect coverage gaps. The process then generates and validates new vetted reference content using the same right vs wrong examples and decision table guardrails.