Skill Forge

Problem

SKILL.md files often fail in practice: instructions are vague, triggers are weak, and script dependencies are inconsistent, making agent execution hard to reproduce. Skill Forge turns Skill authoring into an engineering workflow: requirements capture, architecture selection, file generation, and quality audit.

How It Works

It first classifies intent:
- Create mode: collects the core function, target users, inputs, and outputs, then selects an architecture such as workflow, tool, analysis, generation, aggregation, management, or conversation.
- Audit mode: scores structure, instruction quality, triggers, script quality, documentation, and best practices, then lists actionable fixes.
- Inspiration mode: suggests candidate Skills based on installed skills, recurring pain points, and popular directions.

Generation can produce SKILL.md, scripts, references/ documents, and README.md. Scripts favor Python standard libraries and require UTF-8 encoding, --help, error handling, and emoji-free console output to reduce Windows encoding issues.

Boundaries

It is useful for building agent-facing skill packages with clear steps, tool calls, or document outputs. It is not a substitute for backend development, database design, or complex service architecture. For one-line requests, it ships a first draft and marks uncertain assumptions with [TODO], so key business rules still need human review.