--- title: A Look Inside Our Auto-Reliability Agent Pipeline description: Mabl's engineering team built an auto-reliability pipeline that automatically investigates, fixes, and tests production failures in their AI test authoring... canonical: https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: A Look Inside Our Auto-Reliability Agent Pipeline | daily.dev og:description: Mabl's engineering team built an auto-reliability pipeline that automatically investigates, fixes, and tests production failures in their AI test authoring... og:url: https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks og:image: https://api.daily.dev/og/posts/yh6HyCLKS.png og:image:alt: A Look Inside Our Auto-Reliability Agent Pipeline 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. # A Look Inside Our Auto-Reliability Agent Pipeline **[Mabl Engineering Blog](https://daily.dev/sources/mabl)** · 10 min read · 0 upvotes · 0 comments ## Summary Mabl's engineering team built an auto-reliability pipeline that automatically investigates, fixes, and tests production failures in their AI test authoring system. The pipeline chains a Session Review Agent (SRA) — which has no code access and excels at symptom detection — with Claude-powered investigator and fix-worker agents that verify root causes against the actual codebase before writing any code. Key design decisions include an adversarial investigation phase that debunks incorrect SRA hypotheses, a six-lens simulated peer review panel, and a pre-merge headless execution step. Hard-won lessons include the need for explicit liveness rules to prevent agents from idling, convergence policies to stop infinite refinement loops, and isolating test suite execution to CI rather than the orchestration host. The system reduced human touchpoints to three: batch kickoff, architectural policy calls, and final PR merge. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why does an AI agent that reviews test failure transcripts get the root cause wrong even when it correctly identifies the symptom? An agent with no access to source code can accurately report what happened (error strings, failing tools, message indices) at near 100% accuracy because that's directly observable in the transcript, but its theory of why it happened is only 75-80% accurate because it can only guess at underlying code behavior it cannot see. Mabl's Session Review Agent exhibits exactly this pattern, so its causal hypotheses are treated as unverified leads rather than facts. _Anyone building agent pipelines that diagnose failures can compare architecture patterns like this one on daily.dev._ ### How can you stop an AI code review agent from endlessly iterating on the same pull request? Introduce a strict convergence policy that caps refinement rounds on the same theme, for example allowing three rounds before forcing the system to weigh engineering risk against the threat model, document remaining technical debt, and stop. Without such a rule, a relentless automated reviewer can keep finding technically valid but increasingly impractical critiques, as happened when one review ran nine consecutive rounds on a single PR. _Teams tuning autonomous review loops can find similar operational guardrails on daily.dev._ ### What causes a host machine to run out of memory when running autonomous coding agents in parallel? Running multiple autonomous fix-worker agents that each spin up their own local test suites simultaneously on the same host machine can exhaust available RAM and crash the system, as happened on a first overnight run with four parallel fix workers. The fix is to isolate all functional verification in CI rather than on the orchestration host, while keeping pipeline state recoverable from pushed branches and open PRs. _Developers scaling multi-agent pipelines can track this kind of infrastructure lesson on daily.dev._ ## Similar posts on daily.dev - [How we built an autonomous SRE agent for Kubernetes](https://daily.dev/posts/how-we-built-an-autonomous-sre-agent-for-kubernetes-gxfkkntbg) · LangChain · 6 upvotes · 0 comments - [How We Ship Quality at mabl](https://daily.dev/posts/how-we-ship-quality-at-mabl-yx9gy8bal) · Mabl Engineering Blog · 0 upvotes · 0 comments - [Agents have made CI the bottleneck. Faster pipelines are the wrong fix.](https://daily.dev/posts/agents-have-made-ci-the-bottleneck-faster-pipelines-are-the-wrong-fix--fott4jdmy) · The New Stack · 2 upvotes · 0 comments - [Why Reliability Guardrails Are Needed in Every AI Coding Pipeline](https://daily.dev/posts/why-reliability-guardrails-are-needed-in-every-ai-coding-pipeline-rx8jusfru) · DevOps.com · 1 upvotes · 0 comments --- Tags: [#testing](https://daily.dev/tags/testing), [#ai-agents](https://daily.dev/tags/ai-agents), [#cicd](https://daily.dev/tags/cicd), [#claude](https://daily.dev/tags/claude) [View this post on daily.dev](https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks) ```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":"A Look Inside Our Auto-Reliability Agent Pipeline","url":"https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks"},"datePublished":"2026-07-31T20:00:12.808Z","dateModified":"2026-09-14T07:40:52.788Z","description":"Mabl's engineering team built an auto-reliability pipeline that automatically investigates, fixes, and tests production failures in their AI test authoring...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/479e60287dfdf6277b4769ff5bbe68e3?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/479e60287dfdf6277b4769ff5bbe68e3?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Mabl Engineering Blog","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":"Mabl Engineering Blog","logo":"https://media.daily.dev/image/upload/s--LPBlphgx--/f_auto,q_auto/v1780213707/logos/mabl?_a=BAMAMiWQ0","url":"https://daily.dev/sources/mabl"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"testing,ai-agents,cicd,claude","timeRequired":"PT10M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Mabl Engineering Blog","item":"https://daily.dev/sources/mabl"},{"@type":"ListItem","position":3,"name":"A Look Inside Our Auto-Reliability Agent Pipeline"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/a-look-inside-our-auto-reliability-agent-pipeline-yh6hyclks#faq","mainEntity":[{"@type":"Question","name":"Why does an AI agent that reviews test failure transcripts get the root cause wrong even when it correctly identifies the symptom?","acceptedAnswer":{"@type":"Answer","text":"An agent with no access to source code can accurately report what happened (error strings, failing tools, message indices) at near 100% accuracy because that's directly observable in the transcript, but its theory of why it happened is only 75-80% accurate because it can only guess at underlying code behavior it cannot see. Mabl's Session Review Agent exhibits exactly this pattern, so its causal hypotheses are treated as unverified leads rather than facts. Anyone building agent pipelines that diagnose failures can compare architecture patterns like this one on daily.dev."}},{"@type":"Question","name":"How can you stop an AI code review agent from endlessly iterating on the same pull request?","acceptedAnswer":{"@type":"Answer","text":"Introduce a strict convergence policy that caps refinement rounds on the same theme, for example allowing three rounds before forcing the system to weigh engineering risk against the threat model, document remaining technical debt, and stop. Without such a rule, a relentless automated reviewer can keep finding technically valid but increasingly impractical critiques, as happened when one review ran nine consecutive rounds on a single PR. Teams tuning autonomous review loops can find similar operational guardrails on daily.dev."}},{"@type":"Question","name":"What causes a host machine to run out of memory when running autonomous coding agents in parallel?","acceptedAnswer":{"@type":"Answer","text":"Running multiple autonomous fix-worker agents that each spin up their own local test suites simultaneously on the same host machine can exhaust available RAM and crash the system, as happened on a first overnight run with four parallel fix workers. The fix is to isolate all functional verification in CI rather than on the orchestration host, while keeping pipeline state recoverable from pushed branches and open PRs. Developers scaling multi-agent pipelines can track this kind of infrastructure lesson on daily.dev."}}]} ```