--- title: Ponytail Skill for Claude Code: Does It Really Cut Tokens description: JetBrains ran a rigorous 80-task paired A/B benchmark of the Ponytail skill for Claude Code, which claims to reduce AI-generated code by 54%, tokens by 22%,... canonical: https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Ponytail Skill for Claude Code: Does It Really Cut Tokens | daily.dev og:description: JetBrains ran a rigorous 80-task paired A/B benchmark of the Ponytail skill for Claude Code, which claims to reduce AI-generated code by 54%, tokens by 22%,... og:url: https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt og:image: https://api.daily.dev/og/posts/cG3wamCRt.png og:image:alt: Ponytail Skill for Claude Code: Does It Really Cut Tokens 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. # Ponytail Skill for Claude Code: Does It Really Cut Tokens **[JetBrains](https://daily.dev/sources/jetbrains)** · 19 min read · 1 upvotes · 0 comments ## Summary JetBrains ran a rigorous 80-task paired A/B benchmark of the Ponytail skill for Claude Code, which claims to reduce AI-generated code by 54%, tokens by 22%, cost by 20%, and time by 27%. The measured results: −15% code written, −10.3% cost (statistically significant at p=0.004), and −10.6% time. No quality degradation was detected. Key findings include: the skill never self-activates without its SessionStart hook injection; savings concentrate on larger builds (−31% code on 300+ line tasks, near zero on minimal tasks); and small sample sizes (10 tasks) produced wildly misleading results. Ponytail is the first tool in this series to show a genuine, statistically solid cost saving, though advertised figures are roughly 2–4× larger than what was measured on a general benchmark. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much does the ponytail skill actually reduce code and cost when used with Claude Code, compared to its advertised numbers? Across an 80-task paired benchmark, ponytail cut code written by a median of 15.4% (versus an advertised 54%) and cut cost by 10.3% (versus an advertised 20%), with time down 10.6%. The cost result was statistically solid at p=0.004. Savings were largest on big builds (up to -31%) and near zero on tasks that were already minimal, since the skill works by preventing over-building rather than shrinking necessary code. _Weighing whether a token-saving skill is worth adopting is easier with independently measured numbers like these on daily.dev._ ### Does the ponytail skill for Claude Code self-activate automatically once installed? No. Across ten test sessions where the skill was installed as a plain SKILL.md and left for the model to decide when to use, it self-activated zero times. Ponytail only produces measurable effects when installed as a plugin with its SessionStart hook, which force-injects the ruleset into every session regardless of whether the model would have chosen to use it. _Anyone configuring Claude Code skills can avoid wasted setup by checking activation behavior details like this on daily.dev._ ### Does making Claude Code write less code with a minimalism skill like ponytail hurt code quality? No statistically significant quality difference was detected across 80 paired tasks: 65 scored identically, 6 scored better, and 9 scored slightly worse with ponytail enabled. This is a null result rather than proof of equivalence, since the sample size wasn't powered to rule out small degradations, but it found none of the obvious failure mode where writing less code breaks passing tests. _Developers deciding whether a code-minimizing agent skill is safe to adopt can track findings like this on daily.dev._ ## Similar posts on daily.dev - [A deep dive into Ponytail](https://daily.dev/posts/a-deep-dive-into-ponytail-uztnotcwb) · Flavio Copes · 1 upvotes · 0 comments - [Ponytail Agent Skill Corrects Its Own Benchmark After Contributor Challenge](https://daily.dev/posts/ponytail-agent-skill-corrects-its-own-benchmark-after-contributor-challenge-inn3eqyl5) · InfoQ · 0 upvotes · 0 comments - [Cutting the Coding Agent Bill: 5 Plugins Under Test](https://daily.dev/posts/cutting-the-coding-agent-bill-5-plugins-under-test-hei3opdmj) · marmelab · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude-code](https://daily.dev/tags/claude-code), [#context-engineering](https://daily.dev/tags/context-engineering) [View this post on daily.dev](https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt) ```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":"Ponytail Skill for Claude Code: Does It Really Cut Tokens","url":"https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt"},"datePublished":"2026-07-28T13:39:28.929Z","dateModified":"2026-09-14T06:51:28.366Z","description":"JetBrains ran a rigorous 80-task paired A/B benchmark of the Ponytail skill for Claude Code, which claims to reduce AI-generated code by 54%, tokens by 22%,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/4d6f76b37afb4256023a79ab799ce980?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/4d6f76b37afb4256023a79ab799ce980?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"JetBrains","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":"JetBrains","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/53ecf0c676f34b3896ee109609d91efa","url":"https://daily.dev/sources/jetbrains"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,claude-code,context-engineering","timeRequired":"PT19M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"JetBrains","item":"https://daily.dev/sources/jetbrains"},{"@type":"ListItem","position":3,"name":"Ponytail Skill for Claude Code: Does It Really Cut Tokens"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ponytail-skill-for-claude-code-does-it-really-cut-tokens-cg3wamcrt#faq","mainEntity":[{"@type":"Question","name":"How much does the ponytail skill actually reduce code and cost when used with Claude Code, compared to its advertised numbers?","acceptedAnswer":{"@type":"Answer","text":"Across an 80-task paired benchmark, ponytail cut code written by a median of 15.4% (versus an advertised 54%) and cut cost by 10.3% (versus an advertised 20%), with time down 10.6%. The cost result was statistically solid at p=0.004. Savings were largest on big builds (up to -31%) and near zero on tasks that were already minimal, since the skill works by preventing over-building rather than shrinking necessary code. Weighing whether a token-saving skill is worth adopting is easier with independently measured numbers like these on daily.dev."}},{"@type":"Question","name":"Does the ponytail skill for Claude Code self-activate automatically once installed?","acceptedAnswer":{"@type":"Answer","text":"No. Across ten test sessions where the skill was installed as a plain SKILL.md and left for the model to decide when to use, it self-activated zero times. Ponytail only produces measurable effects when installed as a plugin with its SessionStart hook, which force-injects the ruleset into every session regardless of whether the model would have chosen to use it. Anyone configuring Claude Code skills can avoid wasted setup by checking activation behavior details like this on daily.dev."}},{"@type":"Question","name":"Does making Claude Code write less code with a minimalism skill like ponytail hurt code quality?","acceptedAnswer":{"@type":"Answer","text":"No statistically significant quality difference was detected across 80 paired tasks: 65 scored identically, 6 scored better, and 9 scored slightly worse with ponytail enabled. This is a null result rather than proof of equivalence, since the sample size wasn't powered to rule out small degradations, but it found none of the obvious failure mode where writing less code breaks passing tests. Developers deciding whether a code-minimizing agent skill is safe to adopt can track findings like this on daily.dev."}}]} ```