--- title: Changing One Prompt Can Affect 50 Others — I Built a... description: A pure Python tool models prompt components as versioned collections of named sections and builds a dependency graph to answer, when a shared prompt section... canonical: https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Changing One Prompt Can Affect 50 Others — I Built a Prompt Dependency Graph to Find What Needs Retesting | daily.dev og:description: A pure Python tool models prompt components as versioned collections of named sections and builds a dependency graph to answer, when a shared prompt section... og:url: https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j og:image: https://api.daily.dev/og/posts/lgR0rTu2j.png og:image:alt: Changing One Prompt Can Affect 50 Others — I Built a Prompt Dependency Graph to Find What Needs Retesting 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. # Changing One Prompt Can Affect 50 Others — I Built a Prompt Dependency Graph to Find What Needs Retesting **[Towards Data Science](https://daily.dev/sources/tds)** · 20 min read · 1 upvotes · 0 comments ## Summary A pure Python tool models prompt components as versioned collections of named sections and builds a dependency graph to answer, when a shared prompt section changes, what actually needs re-evaluation. It distinguishes 'Reachable' (everything structurally downstream of a changed component) from the narrower 'Candidate' set (only nodes that declared a dependency on the specific changed section, plus their downstream consumers). Tested on a deterministic 55-node synthetic system with 50 agents and 5 workflows, narrowing ranged from 0% (when a component like tone is used by literally every agent) to 85% (when a component is shared by only ~10% of agents). The relationship between sharing level and narrowing opportunity is monotonic: the more universally shared a component, the less benefit from section-aware tracking. The tool only tells you what to check, not what will actually break behaviorally, and v1 deliberately omits dependency-kind weighting, rename inference, and semantic diffing. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I figure out which prompts need retesting after changing a shared prompt component? Build a dependency graph that tracks which prompt sections each agent declares a dependency on, then compute two sets when a section changes: the full 'reachable' set of everything structurally downstream of the component, and the narrower 'candidate' set of only nodes that declared a dependency on the specific section that changed, plus everything downstream of them. In one 55-node test system, this narrowed a base-policy refund change from 45 reachable nodes down to 24 actual candidates. _daily.dev surfaces practical approaches like dependency-aware evaluation for teams managing composable prompt systems._ ### Why does checking only direct imports of a changed prompt component undercount the actual impact? Checking only direct imports misses transitive dependencies several layers deep, systematically undercounting impact in layered prompt systems. In a test comparing a flat structure (one component feeding 4 agents directly) versus a deep structure (one component feeding through several chained layers to reach the same 4 agents), a one-hop lookup found all 4 impacted nodes in the flat case but only 1 in the deep case, missing three-quarters of the real impact. _daily.dev helps developers dig into dependency analysis techniques before they get burned by an untested downstream break._ ### Does the amount of shared usage of a prompt component affect how much test narrowing is possible? Yes, narrowing potential drops as sharing increases, following a monotonic relationship. In one experiment, a component shared by about 10% of 54 agents allowed 85% narrowing of the evaluation set, while a component shared by 24% allowed 72%, 50% allowed 46%, 76% allowed 20%, and a component used by 100% of agents (tone) allowed 0% narrowing, since every agent is a legitimate candidate. _daily.dev keeps engineers weighing trade-offs like this in view when deciding how to structure shared prompt logic._ ## Similar posts on daily.dev - [Onboarding Agent from Scratch. Part 4: Prompts You Can Test, Not Just Tweak](https://daily.dev/posts/onboarding-agent-from-scratch-part-4-prompts-you-can-test-not-just-tweak-t5upqegtx) · Medium · 0 upvotes · 0 comments --- Tags: [#python](https://daily.dev/tags/python), [#llm](https://daily.dev/tags/llm), [#testing](https://daily.dev/tags/testing), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j) ```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":"Changing One Prompt Can Affect 50 Others — I Built a Prompt Dependency Graph to Find What Needs Retesting","url":"https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j"},"datePublished":"2026-09-03T12:53:19.869Z","dateModified":"2026-09-03T13:43:37.969Z","description":"A pure Python tool models prompt components as versioned collections of named sections and builds a dependency graph to answer, when a shared prompt section...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/e735f18d2554c103d529d49e39e97227?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/e735f18d2554c103d529d49e39e97227?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Towards Data Science","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":"Towards Data Science","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tds","url":"https://daily.dev/sources/tds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"python,llm,testing,prompt-engineering","timeRequired":"PT20M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Towards Data Science","item":"https://daily.dev/sources/tds"},{"@type":"ListItem","position":3,"name":"Changing One Prompt Can Affect 50 Others — I Built a Prompt Dependency Graph to Find What Needs Retesting"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/changing-one-prompt-can-affect-50-others-i-built-a-prompt-dependency-graph-to-find-what-needs-rete-lgr0rtu2j#faq","mainEntity":[{"@type":"Question","name":"How do I figure out which prompts need retesting after changing a shared prompt component?","acceptedAnswer":{"@type":"Answer","text":"Build a dependency graph that tracks which prompt sections each agent declares a dependency on, then compute two sets when a section changes: the full 'reachable' set of everything structurally downstream of the component, and the narrower 'candidate' set of only nodes that declared a dependency on the specific section that changed, plus everything downstream of them. In one 55-node test system, this narrowed a base-policy refund change from 45 reachable nodes down to 24 actual candidates. daily.dev surfaces practical approaches like dependency-aware evaluation for teams managing composable prompt systems."}},{"@type":"Question","name":"Why does checking only direct imports of a changed prompt component undercount the actual impact?","acceptedAnswer":{"@type":"Answer","text":"Checking only direct imports misses transitive dependencies several layers deep, systematically undercounting impact in layered prompt systems. In a test comparing a flat structure (one component feeding 4 agents directly) versus a deep structure (one component feeding through several chained layers to reach the same 4 agents), a one-hop lookup found all 4 impacted nodes in the flat case but only 1 in the deep case, missing three-quarters of the real impact. daily.dev helps developers dig into dependency analysis techniques before they get burned by an untested downstream break."}},{"@type":"Question","name":"Does the amount of shared usage of a prompt component affect how much test narrowing is possible?","acceptedAnswer":{"@type":"Answer","text":"Yes, narrowing potential drops as sharing increases, following a monotonic relationship. In one experiment, a component shared by about 10% of 54 agents allowed 85% narrowing of the evaluation set, while a component shared by 24% allowed 72%, 50% allowed 46%, 76% allowed 20%, and a component used by 100% of agents (tone) allowed 0% narrowing, since every agent is a legitimate candidate. daily.dev keeps engineers weighing trade-offs like this in view when deciding how to structure shared prompt logic."}}]} ```