--- title: Agents, codebases, and teams — Aditya Khandelwal, Amazon... description: A talk from an Amazon AGI Lab engineer who led a 10-person team through adopting AI coding agents in a production codebase. The core argument is that making... canonical: https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab | daily.dev og:description: A talk from an Amazon AGI Lab engineer who led a 10-person team through adopting AI coding agents in a production codebase. The core argument is that making... og:url: https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao og:image: https://api.daily.dev/og/posts/mxAVtl9ao.png og:image:alt: Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 16 min read · 1 upvotes · 0 comments ## Summary A talk from an Amazon AGI Lab engineer who led a 10-person team through adopting AI coding agents in a production codebase. The core argument is that making agents work at team scale is a leadership problem, not an individual one. Key practices shared include: progressive disclosure in codebase structure (keeping CLAUDE.md files under 100 lines and using comments to guide agents to relevant context), building a 'ship it' skill that automates the full code-to-PR-ready pipeline, wiring issues and CI/CD into the repo for self-healing, and continuously iterating on the shared setup. Common failure signals include babysitting agents, silently burning context, and long sessions needing constant intervention. The talk emphasizes winning over skeptics, managing team fear around AI adoption, and treating the agent harness as ongoing organizational infrastructure rather than a one-time setup. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How long should a skill.md file be when setting up Claude Code or similar coding agents for a team? Skill.md files should be kept to a hard limit of around 100 lines, since a skill is really a folder rather than a single monolithic file. The agent's md file should act as a thin index pointing to other files rather than containing everything, because that index is what gets loaded into the first prompt and consumes context immediately. _Teams tuning agent context budgets can track practical setup patterns like this on daily.dev._ ### What are the warning signs that a team's AI coding agent setup isn't working well? Warning signs include engineers babysitting their agents, silently burning through hundreds of thousands of tokens of context and hitting auto-compact on simple tasks, long sessions requiring constant intervention, a growing backlog of low-quality PRs (a 'slop factory'), and blaming the model for being 'dumb' rather than questioning the codebase setup itself. _Anyone debugging flaky agent workflows can compare notes on these failure patterns via daily.dev._ ### What is the 'ship it' skill approach for automating pull requests with a coding agent? It is a single high-value shared skill that takes code from 'done' all the way to 'PR ready for review' automatically: opening the PR, resolving opinions, handling review comments and PR descriptions, writing merge messages, and fixing CI failures, sometimes running for over an hour. Investing team trust in one such skill helps skeptical engineers stop babysitting agents once they see it deliver reliably. _Teams weighing how much PR work to automate with agents can find setup ideas like this on daily.dev._ ## Similar posts on daily.dev - [Inside Kilo Speed: The Engineer Who Teaches Teams How to Think in Agents](https://daily.dev/posts/inside-kilo-speed-the-engineer-who-teaches-teams-how-to-think-in-agents-qsoa3qm4s) · Kilo Blog · 1 upvotes · 0 comments - [AWS says almost 90% of Amazon’s early AI agent prototypes never shipped](https://daily.dev/posts/aws-says-almost-90-of-amazon-s-early-ai-agent-prototypes-never-shipped-nz5oy4r9c) · The Next Web · 1 upvotes · 0 comments - [How frontier teams are reinventing AI-native development](https://daily.dev/posts/how-frontier-teams-are-reinventing-ai-native-development-fm1xezahj) · AWS · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#productivity](https://daily.dev/tags/productivity), [#cicd](https://daily.dev/tags/cicd), [#leadership](https://daily.dev/tags/leadership), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab","url":"https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao"},"datePublished":"2026-08-11T00:16:33.617Z","dateModified":"2026-09-13T19:19:10.662Z","description":"A talk from an Amazon AGI Lab engineer who led a 10-person team through adopting AI coding agents in a production codebase. The core argument is that making...","image":"https://i.ytimg.com/vi/aeTb5BdmTTc/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/aeTb5BdmTTc/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,productivity,cicd,leadership,prompt-engineering","timeRequired":"PT16M","video":{"@type":"VideoObject","name":"Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab","description":"A talk from an Amazon AGI Lab engineer who led a 10-person team through adopting AI coding agents in a production codebase. The core argument is that making...","thumbnailUrl":"https://i.ytimg.com/vi/aeTb5BdmTTc/sddefault.jpg","uploadDate":"2026-08-11T00:16:33.617Z","duration":"PT16M","url":"https://api.daily.dev/r/mxAVtl9ao","embedUrl":"https://www.youtube.com/embed/aeTb5BdmTTc"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Agents, codebases, and teams — Aditya Khandelwal, Amazon AGI Lab"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/agents-codebases-and-teams-aditya-khandelwal-amazon-agi-lab-mxavtl9ao#faq","mainEntity":[{"@type":"Question","name":"How long should a skill.md file be when setting up Claude Code or similar coding agents for a team?","acceptedAnswer":{"@type":"Answer","text":"Skill.md files should be kept to a hard limit of around 100 lines, since a skill is really a folder rather than a single monolithic file. The agent's md file should act as a thin index pointing to other files rather than containing everything, because that index is what gets loaded into the first prompt and consumes context immediately. Teams tuning agent context budgets can track practical setup patterns like this on daily.dev."}},{"@type":"Question","name":"What are the warning signs that a team's AI coding agent setup isn't working well?","acceptedAnswer":{"@type":"Answer","text":"Warning signs include engineers babysitting their agents, silently burning through hundreds of thousands of tokens of context and hitting auto-compact on simple tasks, long sessions requiring constant intervention, a growing backlog of low-quality PRs (a 'slop factory'), and blaming the model for being 'dumb' rather than questioning the codebase setup itself. Anyone debugging flaky agent workflows can compare notes on these failure patterns via daily.dev."}},{"@type":"Question","name":"What is the 'ship it' skill approach for automating pull requests with a coding agent?","acceptedAnswer":{"@type":"Answer","text":"It is a single high-value shared skill that takes code from 'done' all the way to 'PR ready for review' automatically: opening the PR, resolving opinions, handling review comments and PR descriptions, writing merge messages, and fixing CI failures, sometimes running for over an hour. Investing team trust in one such skill helps skeptical engineers stop babysitting agents once they see it deliver reliably. Teams weighing how much PR work to automate with agents can find setup ideas like this on daily.dev."}}]} ```