StarZero All shows Try the podcast clipper →

Aug 19, 2026 · 32m · startup-ideas

How I Use Skills + AI Agents to Run My Life

Greg Isenberg · 6m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Greg Isenberg hosts Remy from 'AI with Remy' to break down how businesses can transition from isolated AI prompts to collaborative, multiplayer agent workflows. They explore managing AI skills as markdown Standard Operating Procedures, distributing them across teams via centralized GitHub plugin repositories, and chaining modular skills to automate complex business operations.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 23% of the talking time here. How this is scored →

Greg as informed peer 4.7 Guest teaching 6.2 Guest disagreement 0.1 Greg pushing back 0.2
05100:0010:0020:0030:001:39–5:13 · Greg as informed peer 4/10 Welcoming Remy and Setting Episode Objectives Greg sets the stage and prompts Remy to explain agent skills. Remy breaks down what skills are (SOPs for AI) and demonstrates how they eliminate repetitive prompt formatting and prevent output slop.5:13–8:07 · Greg as informed peer 3/10 The Single-Player AI Bottleneck and Real-World Skills Remy uses the Microsoft Word era analogy to explain how skills are currently stuck in single-player mode on individual machines, detailing his custom Notion, brand voice, and Resend formatting skills.8:07–12:00 · Greg as informed peer 5/10 Flaws in Traditional Skill Distribution Methods Remy explains the technical pitfalls of sharing skill folders via Google Drive, Dropbox, or Obsidian symlinks, advocating for GitHub repositories as plugins. Greg reassures non-technical listeners that repo setups are straightforward.12:00–14:58 · Greg as informed peer 5/10 Live Demonstration of Plugin Installation in Claude and Codex Remy demonstrates installing custom skill bundles as departmental plugins inside Claude Code terminal and Codex. Greg confirms cross-harness compatibility across tools like ChatGPT and Codex.14:59–20:02 · Greg as informed peer 5/10 Plugin Mechanics, Version Control, and Pre-Built Skills Greg asks why Remy maintains a separate personal repo if skills aren't shared, prompting Remy's cautionary tale of Claude deleting 150 unbacked skills. Greg validates pre-built creator skills from creators like Corey Haines.20:02–24:12 · Greg as informed peer 4/10 Building a Visual Dashboard, AI Assistant, and Usage Analytics Remy showcases his custom web dashboard frontend connected to GitHub, featuring visual skill chains, an AI skills query assistant, and usage analytics hooks. Greg quips about tracking neglected skills.24:14–28:22 · Greg as informed peer 5/10 Skill Chaining Frameworks and Self-Improvement Loops Greg inquires about frameworks for determining skill granularity. Remy explains decomposing complex processes like YouTube publishing into modular chains and demonstrates adding automated self-improvement loops at the bottom of markdown files.28:23–30:44 · Greg as informed peer 5/10 The 'Thin Agents, Thick Skills' Philosophy in Action Remy details the 'thin agents, thick skills' architecture and shares a concrete example of a teammate fixing an image bug in Resend that propagated across the team. Greg synthesizes the strategic benefits of deterministic skill execution.30:44–32:26 · Greg as informed peer 6/10 The Era of Multiplayer AI and Final Thoughts Greg connects the evolution of single-player to multiplayer AI to classic software trends like Photoshop leading into Figma, offering a sharp conceptual synthesis to close the interview.1:39–5:13 · Guest teaching 6/10 Welcoming Remy and Setting Episode Objectives Greg sets the stage and prompts Remy to explain agent skills. Remy breaks down what skills are (SOPs for AI) and demonstrates how they eliminate repetitive prompt formatting and prevent output slop.5:13–8:07 · Guest teaching 7/10 The Single-Player AI Bottleneck and Real-World Skills Remy uses the Microsoft Word era analogy to explain how skills are currently stuck in single-player mode on individual machines, detailing his custom Notion, brand voice, and Resend formatting skills.8:07–12:00 · Guest teaching 7/10 Flaws in Traditional Skill Distribution Methods Remy explains the technical pitfalls of sharing skill folders via Google Drive, Dropbox, or Obsidian symlinks, advocating for GitHub repositories as plugins. Greg reassures non-technical listeners that repo setups are straightforward.12:00–14:58 · Guest teaching 6/10 Live Demonstration of Plugin Installation in Claude and Codex Remy demonstrates installing custom skill bundles as departmental plugins inside Claude Code terminal and Codex. Greg confirms cross-harness compatibility across tools like ChatGPT and Codex.14:59–20:02 · Guest teaching 6/10 Plugin Mechanics, Version Control, and Pre-Built Skills Greg asks why Remy maintains a separate personal repo if skills aren't shared, prompting Remy's cautionary tale of Claude deleting 150 unbacked skills. Greg validates pre-built creator skills from creators like Corey Haines.20:02–24:12 · Guest teaching 7/10 Building a Visual Dashboard, AI Assistant, and Usage Analytics Remy showcases his custom web dashboard frontend connected to GitHub, featuring visual skill chains, an AI skills query assistant, and usage analytics hooks. Greg quips about tracking neglected skills.24:14–28:22 · Guest teaching 7/10 Skill Chaining Frameworks and Self-Improvement Loops Greg inquires about frameworks for determining skill granularity. Remy explains decomposing complex processes like YouTube publishing into modular chains and demonstrates adding automated self-improvement loops at the bottom of markdown files.28:23–30:44 · Guest teaching 6/10 The 'Thin Agents, Thick Skills' Philosophy in Action Remy details the 'thin agents, thick skills' architecture and shares a concrete example of a teammate fixing an image bug in Resend that propagated across the team. Greg synthesizes the strategic benefits of deterministic skill execution.30:44–32:26 · Guest teaching 4/10 The Era of Multiplayer AI and Final Thoughts Greg connects the evolution of single-player to multiplayer AI to classic software trends like Photoshop leading into Figma, offering a sharp conceptual synthesis to close the interview.1:39–5:13 · Guest disagreement 0/10 Welcoming Remy and Setting Episode Objectives Greg sets the stage and prompts Remy to explain agent skills. Remy breaks down what skills are (SOPs for AI) and demonstrates how they eliminate repetitive prompt formatting and prevent output slop.5:13–8:07 · Guest disagreement 0/10 The Single-Player AI Bottleneck and Real-World Skills Remy uses the Microsoft Word era analogy to explain how skills are currently stuck in single-player mode on individual machines, detailing his custom Notion, brand voice, and Resend formatting skills.8:07–12:00 · Guest disagreement 1/10 Flaws in Traditional Skill Distribution Methods Remy explains the technical pitfalls of sharing skill folders via Google Drive, Dropbox, or Obsidian symlinks, advocating for GitHub repositories as plugins. Greg reassures non-technical listeners that repo setups are straightforward.12:00–14:58 · Guest disagreement 0/10 Live Demonstration of Plugin Installation in Claude and Codex Remy demonstrates installing custom skill bundles as departmental plugins inside Claude Code terminal and Codex. Greg confirms cross-harness compatibility across tools like ChatGPT and Codex.14:59–20:02 · Guest disagreement 0/10 Plugin Mechanics, Version Control, and Pre-Built Skills Greg asks why Remy maintains a separate personal repo if skills aren't shared, prompting Remy's cautionary tale of Claude deleting 150 unbacked skills. Greg validates pre-built creator skills from creators like Corey Haines.20:02–24:12 · Guest disagreement 0/10 Building a Visual Dashboard, AI Assistant, and Usage Analytics Remy showcases his custom web dashboard frontend connected to GitHub, featuring visual skill chains, an AI skills query assistant, and usage analytics hooks. Greg quips about tracking neglected skills.24:14–28:22 · Guest disagreement 0/10 Skill Chaining Frameworks and Self-Improvement Loops Greg inquires about frameworks for determining skill granularity. Remy explains decomposing complex processes like YouTube publishing into modular chains and demonstrates adding automated self-improvement loops at the bottom of markdown files.28:23–30:44 · Guest disagreement 0/10 The 'Thin Agents, Thick Skills' Philosophy in Action Remy details the 'thin agents, thick skills' architecture and shares a concrete example of a teammate fixing an image bug in Resend that propagated across the team. Greg synthesizes the strategic benefits of deterministic skill execution.30:44–32:26 · Guest disagreement 0/10 The Era of Multiplayer AI and Final Thoughts Greg connects the evolution of single-player to multiplayer AI to classic software trends like Photoshop leading into Figma, offering a sharp conceptual synthesis to close the interview.1:39–5:13 · Greg pushing back 0/10 Welcoming Remy and Setting Episode Objectives Greg sets the stage and prompts Remy to explain agent skills. Remy breaks down what skills are (SOPs for AI) and demonstrates how they eliminate repetitive prompt formatting and prevent output slop.5:13–8:07 · Greg pushing back 0/10 The Single-Player AI Bottleneck and Real-World Skills Remy uses the Microsoft Word era analogy to explain how skills are currently stuck in single-player mode on individual machines, detailing his custom Notion, brand voice, and Resend formatting skills.8:07–12:00 · Greg pushing back 1/10 Flaws in Traditional Skill Distribution Methods Remy explains the technical pitfalls of sharing skill folders via Google Drive, Dropbox, or Obsidian symlinks, advocating for GitHub repositories as plugins. Greg reassures non-technical listeners that repo setups are straightforward.12:00–14:58 · Greg pushing back 0/10 Live Demonstration of Plugin Installation in Claude and Codex Remy demonstrates installing custom skill bundles as departmental plugins inside Claude Code terminal and Codex. Greg confirms cross-harness compatibility across tools like ChatGPT and Codex.14:59–20:02 · Greg pushing back 1/10 Plugin Mechanics, Version Control, and Pre-Built Skills Greg asks why Remy maintains a separate personal repo if skills aren't shared, prompting Remy's cautionary tale of Claude deleting 150 unbacked skills. Greg validates pre-built creator skills from creators like Corey Haines.20:02–24:12 · Greg pushing back 0/10 Building a Visual Dashboard, AI Assistant, and Usage Analytics Remy showcases his custom web dashboard frontend connected to GitHub, featuring visual skill chains, an AI skills query assistant, and usage analytics hooks. Greg quips about tracking neglected skills.24:14–28:22 · Greg pushing back 0/10 Skill Chaining Frameworks and Self-Improvement Loops Greg inquires about frameworks for determining skill granularity. Remy explains decomposing complex processes like YouTube publishing into modular chains and demonstrates adding automated self-improvement loops at the bottom of markdown files.28:23–30:44 · Greg pushing back 0/10 The 'Thin Agents, Thick Skills' Philosophy in Action Remy details the 'thin agents, thick skills' architecture and shares a concrete example of a teammate fixing an image bug in Resend that propagated across the team. Greg synthesizes the strategic benefits of deterministic skill execution.30:44–32:26 · Greg pushing back 0/10 The Era of Multiplayer AI and Final Thoughts Greg connects the evolution of single-player to multiplayer AI to classic software trends like Photoshop leading into Figma, offering a sharp conceptual synthesis to close the interview.

speaking balance: gold is Greg, purple is the guest (3 minute bins)

0:00 · Greg 69.7% · guest 30.3%0:00 · Greg 69.7% · guest 30.3%3:00 · Greg 6.4% · guest 93.6%3:00 · Greg 6.4% · guest 93.6%6:00 · Greg 0% · guest 100%6:00 · Greg 0% · guest 100%9:00 · Greg 7.3% · guest 92.7%9:00 · Greg 7.3% · guest 92.7%12:00 · Greg 5.2% · guest 94.8%12:00 · Greg 5.2% · guest 94.8%15:00 · Greg 22.3% · guest 77.7%15:00 · Greg 22.3% · guest 77.7%18:00 · Greg 27.2% · guest 72.8%18:00 · Greg 27.2% · guest 72.8%21:00 · Greg 0% · guest 100%21:00 · Greg 0% · guest 100%24:00 · Greg 27.7% · guest 72.3%24:00 · Greg 27.7% · guest 72.3%27:00 · Greg 36.3% · guest 63.7%27:00 · Greg 36.3% · guest 63.7%30:00 · Greg 61.5% · guest 38.5%30:00 · Greg 61.5% · guest 38.5%
Sharpest disagreement ▶ 8:30 Pushing Back Against Cloud Drive Workarounds

In a very collaborative interview, Remy's strongest pushback is against conventional file-sharing methods like Google Drive and Obsidian, pointing out that symlink hacks break immediately when onboarding non-technical teammates.

 Hardest push from Greg ▶ 17:57 Questioning Personal Repositories

Greg presses Remy on why anyone would bother maintaining a personal skills repo if they are not actively collaborating or sharing it with teammates.

 Biggest teaching moment ▶ 27:10 Autonomous Self-Improvement Loops in Skill Files

Remy educates Greg on embedding an automatic reflection prompt at the bottom of skill markdown files so the agent proactively proposes structural updates after errors.

 Greg holds their own ▶ 31:12 Multiplayer AI Evolution via Photoshop-Figma Analogy

Greg demonstrates his macro tech knowledge by contextualizing Remy's tactical system within the broader software paradigm of single-player tools inevitably evolving into multiplayer platforms.
the scores for every segment, with the reasoning behind each
ChapterTopicGreg as informed peerGuest teachingGuest disagreementGreg pushing backWhy
Welcoming Remy and Setting Episode Objectives 4600 Greg sets the stage and prompts Remy to explain agent skills. Remy breaks down what skills are (SOPs for AI) and demonstrates how they eliminate repetitive prompt formatting and prevent output slop.
The Single-Player AI Bottleneck and Real-World Skills 3700 Remy uses the Microsoft Word era analogy to explain how skills are currently stuck in single-player mode on individual machines, detailing his custom Notion, brand voice, and Resend formatting skills.
Flaws in Traditional Skill Distribution Methods 5711 Remy explains the technical pitfalls of sharing skill folders via Google Drive, Dropbox, or Obsidian symlinks, advocating for GitHub repositories as plugins. Greg reassures non-technical listeners that repo setups are straightforward.
Live Demonstration of Plugin Installation in Claude and Codex 5600 Remy demonstrates installing custom skill bundles as departmental plugins inside Claude Code terminal and Codex. Greg confirms cross-harness compatibility across tools like ChatGPT and Codex.
Plugin Mechanics, Version Control, and Pre-Built Skills 5601 Greg asks why Remy maintains a separate personal repo if skills aren't shared, prompting Remy's cautionary tale of Claude deleting 150 unbacked skills. Greg validates pre-built creator skills from creators like Corey Haines.
Building a Visual Dashboard, AI Assistant, and Usage Analytics 4700 Remy showcases his custom web dashboard frontend connected to GitHub, featuring visual skill chains, an AI skills query assistant, and usage analytics hooks. Greg quips about tracking neglected skills.
Skill Chaining Frameworks and Self-Improvement Loops 5700 Greg inquires about frameworks for determining skill granularity. Remy explains decomposing complex processes like YouTube publishing into modular chains and demonstrates adding automated self-improvement loops at the bottom of markdown files.
The 'Thin Agents, Thick Skills' Philosophy in Action 5600 Remy details the 'thin agents, thick skills' architecture and shares a concrete example of a teammate fixing an image bug in Resend that propagated across the team. Greg synthesizes the strategic benefits of deterministic skill execution.
The Era of Multiplayer AI and Final Thoughts 6400 Greg connects the evolution of single-player to multiplayer AI to classic software trends like Photoshop leading into Figma, offering a sharp conceptual synthesis to close the interview.

Statements from this episode (2)

Insight
Isenberg: Poor AI output stems from lacking tight skills and SOPs
“A lot of people use AI. And they get slop, and a big reason is they don't have tight skills.”
Greg Isenberg Aug 19, 2026 ▶ 4:08
Insight
Isenberg: Teams can adopt pre-built AI skills rather than starting from scratch
“There's trusted creators and people who have skills on the internet that, you know, you don't need to start from scratch. So you can basically download some of these skills, have that as your foundation, start adding skills based on, you know, The type of busi…”
Greg Isenberg Aug 19, 2026 ▶ 16:35
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more