{"slug": "i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it", "title": "I benchmarked a Scrum workflow for AI coding agents. Here's what it fixed, and what it costs.", "summary": "A developer benchmarked a Scrum-style workflow for teams of Claude Code coding agents, comparing the Lanewright skill against an unaided baseline across six runs per arm on a small two-module Go repository. Runs using the skill completed 6 of 6 versus 3 of 6 without it, with every miss traced to uncommitted reports and sprint summaries, while all 26 merges across 12 runs waited for a green CI run on the PR head. The skill cost about 1.5× more ($5.49 vs $3.55 per sprint) and 1.4× the time, dropping to $3.37 with an orchestrator-on-Sonnet cost setting.", "body_md": "Parallel coding agents are fast until two of them edit the same file. Or until the implementer also writes the tests that are supposed to check it. Or until something merges because \"the tests pass on my machine\".\n\nI've been running sprints with teams of Claude Code agents on real projects for a while, first with a free plugin I wrote, scrum-kit, and now with a rebuilt version, Lanewright. This post is about measuring whether the workflow actually helps, because \"it feels more organised\" is not a reason to pay for extra tokens.\n\nThe workflow\n\nOne session plays Scrum Master. It never writes application code. It hands each backlog item to roles: an implementer, a QA engineer, a reviewer, and where the project needs them a schema owner, an API contract keeper, a tooling owner and a security auditor.\n\nThree rules do most of the work:\n\nEvery role gets its own git worktree, limited to its region of the repo. A config file, .claude/kit.json, says which paths each role may write. In Claude Code, a small Python script turns that into permissions.deny rules inside the worktree, so an implementer started there simply cannot edit the tests.\n\nAn ownership check runs before every commit, on any agent. It lists any changed, staged or untracked file outside the role's region and fails.\n\nNothing merges without evidence. One pull request per item, merged with a merge commit only after a green CI run on that PR's exact head and a review by a session that wrote neither the code nor the tests.\n\nThe benchmark\n\nOne sprint of a small two-module Go repository. Item one adds a function. Item two runs a code generator that also writes into the other module, a file no role is allowed to commit.\n\nBoth arms got the same written Definition of Done: separate implementer, QA and reviewer, one PR per item merged only after green CI on its head, regions respected, reports, and a draft integration PR. The only difference: one arm was told to use the skill. Three runs per arm, in Claude Code (Opus 5.5) and in Codex CLI, scored from git and the GitHub API.\n\nResults\n\nComplete runs: 6 of 6 with the skill, 3 of 6 without. Every miss was the same: reports and the sprint summary were never committed. Two Codex runs said the config had no region for documents. The skill defines one, so nobody has to guess.\n\n26 of 26 merges waited for a green CI run on the PR head, in all 12 runs. When you write the process out, a strong model follows most of it on its own. The skill's value is completeness and consistency, not doing work the model would skip.\n\nCost: in Claude Code with Opus everywhere, $5.49 per sprint with the skill against $3.55 without, about 1.5× the cost and 1.4× the time. With the skill's cost settings (orchestrator on Sonnet 5.5, roles on Opus 5.5), three more runs cost $3.37 on average and still met every item.\n\nLimits\n\nSix runs per side on one small repo, all on Windows. Read the numbers as a direction, not a precise rate. Tool-level file locks exist only in Claude Code; on other agents the ownership check before each commit is the only enforcement, and it reports rather than blocks.\n\nTry it\n\nThe method, prompts, scoring script and per-run data are public: [https://github.com/huzairsaputra/lanewright](https://github.com/huzairsaputra/lanewright)\n\nLanewright itself is a paid skill on Agensi (this is my referral link): [https://www.agensi.io/r/HUZAIRSAPUTRA?to=/skills/lanewright-sprint-workflow-for-ai-coding-agent-teams](https://www.agensi.io/r/HUZAIRSAPUTRA?to=/skills/lanewright-sprint-workflow-for-ai-coding-agent-teams)\n\nIf you'd rather start free, scrum-kit is MIT-licensed on GitHub (its prompts are in Bahasa Indonesia): [https://github.com/huzairsaputra/scrum-kit](https://github.com/huzairsaputra/scrum-kit)", "url": "https://wpnews.pro/news/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it", "canonical_source": "https://dev.to/huzair_saputra/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-heres-what-it-fixed-and-what-it-costs-4aja", "published_at": "2026-11-06 17:46:17+00:00", "updated_at": "2026-11-06 17:53:50.171629+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "mlops"], "entities": ["Lanewright", "scrum-kit", "Claude Code", "Codex CLI", "Agensi", "GitHub", "Anthropic", "OpenAI"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it", "markdown": "https://wpnews.pro/news/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it.md", "text": "https://wpnews.pro/news/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it.txt", "jsonld": "https://wpnews.pro/news/i-benchmarked-a-scrum-workflow-for-ai-coding-agents-here-s-what-it-fixed-and-it.jsonld"}}