| name | cherry-picked-agent-skills |
| description | 6 unique agent skills cherry-picked from Addy Osmani's agent-skills — interview extraction, idea refinement, adversarial review, source-driven development, context engineering, and deprecation workflows. Use when requirements gathering, idea refinement, adversarial review, documentation-driven dev, context optimization, code deprecation. |
| domain | development |
| author | oyi77 |
| license | Apache-2.0 |
| subdomain | software-development |
| tags | ["agent-skills","methodology","interview","context-engineering","deprecation","adversarial-review"] |
| version | 1.0.0 |
Overview
Six unique agent workflow skills cherry-picked from addyosmani/agent-skills that do not overlap with existing 1ai-skills. Each skill defines a structured methodology for AI agents to follow during specific development workflows.
Skill 1: interview-me
Description: One-question-at-a-time interview that extracts what the user actually wants, not what they initially say they want.
When to use: Starting any non-trivial feature or project. When requirements are vague. When the user says "just build X" without clear specifications.
Key Principles:
- Ask one question at a time — never batch questions
- Mirror back what you heard before asking the next question
- Dig into the "why" behind each request
- Surface implicit constraints the user hasn't mentioned
- Stop interviewing when you can state the requirement back and the user confirms
Process:
- Restate the user's initial request in your own words
- Ask: "What problem does this solve for you?"
- Ask about constraints: timeline, tech stack, existing code, audience
- Ask about success criteria: how will you know this is done?
- Summarize the extracted requirements and ask for confirmation
- Only then begin planning or implementation
Skill 2: idea-refine
Description: Divergent then convergent thinking to turn vague ideas into concrete, actionable proposals.
When to use: Brainstorming sessions. When the user has a rough concept but no clear path. When multiple approaches are possible.
Key Principles:
- Divergent phase: generate many options without judgment
- Convergent phase: evaluate, filter, and combine into a single proposal
- Never skip divergent thinking — jumping to solutions misses better alternatives
- Score options on feasibility, impact, and effort
Process:
- Divergent: Generate 5-10 possible interpretations of the idea
- Expand: For each interpretation, list pros, cons, and dependencies
- Cluster: Group related interpretations together
- Score: Rate each cluster on feasibility (1-5), impact (1-5), effort (1-5)
- Convergent: Select top 1-3 candidates, combine best elements
- : Present a concrete proposal with clear next steps
