| name | recipe-upgrader |
| description | Automates the upgrade of blueprint recipes to the "Work-Ready" format. Use when converting old blueprints in content/recipes/ to the new format with archetypes (Processor, Researcher, Hybrid) and simplified prompts. |
Recipe Upgrader
Overview
This skill transforms old, technical blueprints into user-friendly "Work-Ready" versions. It standardizes frontmatter, simplifies sections, and rewrites prompts into a structured three-phase execution logic.
Workflow
1. Identify Archetype
Analyze the existing blueprint's logic to categorize it into one of the three archetypes:
- Processor: Use if it primarily transforms local files (CSV, PDF, etc.).
- Researcher: Use if it primarily scrapes or searches the web.
- Hybrid: Use if it does both or enriches local files with web data.
Refer to archetypes.md for detailed logic patterns.
2. Standardize Frontmatter
- Update the frontmatter block.
- Add
archetype: Processor | Researcher | Hybrid.
- Set
difficulty and time to realistic human values.
- Ensure
sampleData is correctly formatted to seed missing files.
3. Simplify Content
- Delete all technical sections (e.g., "Run this with AI", "Agent Configuration").
- Generate the following user sections in plain English:
# What This Does# What You Need# What You Get# How to Use
Refer to conversion-guide.md for structural rules.
4. Rewrite the Prompt
Rewrite the blueprint's strategy into a high-quality prompt for a modern agent:
- Role & Objective: Start with "You are a [Role]. Your job is to [Goal]."
- Phase 1: Setup: Handle file reading and sample data creation.
- Phase 2: Execution: Loop through items/rows.
- Phase 3: Output: Save results to a structured file (CSV or MD).
- Ending: Always end the prompt with "Start now."
Quality Standards
- No tool-specific jargon (e.g., don't mention
web_fetch or run_shell_command).
- Every recipe must support batch processing if it involves data.
- Output filenames must be clear (e.g.,
results.csv, report.md).
Reference Material
