Program Ideation

Before writing code, search for existing libraries, tools, and frameworks that already solve the problem. AI tools have a bias toward reimplementing things from scratch — this skill counteracts that by front-loading research.

Process

1. Parse the user's requirements into 3-5 distinct search angles. Each angle should target a different facet of what the user wants:

   • The core capability (e.g., "Python library for PDF table extraction")
   • The domain/ecosystem (e.g., "data pipeline tools for converting PDF to CSV")
   • Alternative framings (e.g., "OCR library with table detection")
   • Known adjacent tools (e.g., "Tabula alternatives 2025")
   • The "awesome list" angle (e.g., "awesome-pdf-parsing GitHub")

2. Launch parallel web search agents — one per search angle. Each agent should:

   • Perform 2-3 web searches with different query phrasings
   • For each promising result, fetch the project page or README to get: name, what it does, how mature it is (stars, last commit, maintenance status), and how it compares to alternatives
   • Return a structured summary: library name, URL, one-line description, maturity signals, and how well it fits the stated requirements

3. Synthesize and present results on the main thread:

   • Group findings by approach/category
   • For each candidate, list: name, what it does, fit to requirements, maturity
   • Flag any gaps — requirements that no existing tool covers
   • Recommend a shortlist (1-3 options) with reasoning

4. Iterate with the user — the first search may not match what they actually want. Ask clarifying questions. Re-search with refined angles if needed.

Agent Prompt Template

Use this template when spawning each search agent:

You are researching existing tools/libraries for the following need:

**User's requirement:** {requirement_summary}
**Search angle:** {specific_angle}

Do 2-3 web searches with different phrasings. For each promising project you find:
1. Fetch its homepage or GitHub README
2. Note: name, URL, one-line description, stars/maintenance status, key features
3. Assess: how well does it fit the stated requirement? What does it cover vs. miss?

Return a structured summary of your findings. Prefer recent, actively maintained projects.

When Applying This Skill

1. Do NOT write any implementation code until research is complete and the user has chosen an approach
2. Spawn at least 3 parallel search agents with genuinely different search angles — not minor query variations
3. If the user's description is vague, ask one round of clarifying questions before searching
4. After presenting results, explicitly ask: "Does this match what you're looking for, or should I search from a different angle?"
5. If nothing good exists, say so — then the user can make an informed decision to build from scratch