prompt-factory

I took a look at prompt-factory and noticed you're tackling template composition at scale—the architecture has some solid fundamentals, but there's room to tighten up the implementation patterns and error handling to really unlock its potential.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 63/100, which is solidly in D territory. This is based on Anthropic's best practices for agentic skills. Your strongest area is Utility (16/20)—the preset library and validation gates are genuinely useful. The weakest link is Progressive Disclosure Architecture (18/30)—your SKILL.md is bloated at 1,110 lines when it should be ~250, which kills token efficiency.

What's Working Well

• Comprehensive preset system: 69 presets across 15 domains give real value. The 7-point validation gates are a solid quality check that users will appreciate.
• Multi-format support: Handling XML, JSON, YAML, and plain text output shows you understand diverse use cases.
• Rich discoverability: Your trigger terms are well-chosen—"prompt factory," "mega-prompt," references to Claude, ChatGPT, Gemini all help users find this when they need it.
• Structured approach: The 5-7 question flow creates predictable, repeatable results, which is exactly what users want from a factory pattern.

The Big One: SKILL.md Token Bloat

This is your biggest win waiting to happen. Your SKILL.md is 1,110 lines (~37K tokens) when Progressive Disclosure Architecture demands something closer to 250 lines. You're violating the core principle: overview + references.

Why it matters: Every word in SKILL.md burns tokens for users. Duplicating your full preset list (lines 93-124, 797-919), complete XML templates (lines 373-483), and repeating scope warnings 5+ times creates redundancy that eats context budget.

The fix: Restructure into:

1. Overview (~50 lines): What it does, triggers, quick decision tree
2. Question flow (~40 lines): The 5-7 questions you ask, not the full preset list
3. Format options (~30 lines): Brief summary of output formats
4. Validation gates (~30 lines): The 7-point checks
5. See presets reference (~20 lines): "For 69 presets across 15 domains, see [reference file]"
6. References (~50 lines): Links to best-practices-reference.md where the full preset list lives

Move preset definitions, complete template structures, and detailed examples to a separate references/ file. SKILL.md becomes a map; references become the territory.

Impact: +6 points, cleaner PDA score.

Other Things Worth Fixing

1. Marketing language kills objectivity — "World-class prompt powerhouse" and "production-ready mega-prompts" read like ad copy. Replace with plain instruction: "Generate structured prompts for any role, industry, and task through 7-question flow and 69 presets."

2. Terminology inconsistency creates friction — You alternate between "mega-prompt"/"prompt," "preset"/"template," "questionnaire"/"questions." Pick one term per concept and stick with it consistently. "Prompt" not "mega-prompt," "preset" not "template."

3. Add a table of contents — A 1,110-line file needs a TOC so users can jump to sections. Even after you trim it down, include this at the top.

4. Missing trigger phrases in frontmatter — Your description field should include explicit triggers: "Use when asked to 'prompt factory', 'run prompt factory', or 'generate prompts for a role'." This directly impacts spec compliance scoring.

Quick Wins

• Trim SKILL.md to ~250 lines — Move presets and templates to references. +6 points.
• Swap marketing language for plain instruction — Takes 15 minutes, fixes writing style. +3 points.
• Standardize terminology — One term per concept throughout the file. +2 points.
• Add trigger phrases — Improves spec compliance score. +2 points.

These four fixes alone get you from 63 → ~76, moving you into C+ territory.

Checkout your skill here: SkillzWave.ai | SpillWave We have an agentic skill installer that install skills in 14+ coding agent platforms. Check out this guide on how to improve your agentic skills.