PR Custody

Add research-prompt-builder skill

@kharevanaichecks n/achecks…add-research-prompt-builder → master2 files · +162 −0updated 1w ago

▸Description

What problem it solves

Building a good deep research prompt is harder than it looks. Most prompts sent to Perplexity or ChatGPT miss key angles, have no fallback rule when results are empty, and produce vague outputs with no required format. This skill fixes that.

Who uses this workflow

Engineers and founders evaluating tools before buying, researchers doing competitive analysis, anyone who regularly hands off research tasks to Perplexity, ChatGPT, or similar deep research tools.

What it does

Asks 4 clarifying questions (what to research, what for, time window, depth tier) before generating anything
Writes a full search plan with exact query strings, required output format per finding, and a zero-result fallback rule
Runs an internal 9.5/10 quality gate across 5 dimensions (Specificity, Completeness, Enforceability, Stack-awareness, Edge case coverage) and iterates until threshold is met
Can also run the research itself directly if the user prefers not to export to another tool

Tested on

Claude.ai and Claude Code

Example

User: "Generate a research prompt for evaluating Notion as a team wiki. Last 6 months."

Output: A detailed prompt with 20+ targeted search queries across Reddit, HN, G2, Capterra, YouTube, and switching signal searches, with required output format per source and a summary table at the end. Quality audit appended showing 9.6/10.

loading diff…