Guide · Agent skill
How to use the Ponytail Gain Scoreboard agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The ponytail-gain skill is a small agent skill from the Ponytail project that displays a one-time text scoreboard of published benchmark averages for the Ponytail skill. It compares an agent using the skill against the same agent without it, across lines of code, output tokens, cost, time and pass rates. This guide explains what an agent skill is, how to install this one, how to trigger it, and what its numbers do and do not tell you.
Ponytail Gain ScoreboardDietrichGebert/ponytail · MIT Download the skill Quick start
- 1Download the ZIP. Get the ponytail-gain ZIP from this page and unzip it.
- 2Choose where to install. Place the skill folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Ask for the scoreboard. Make a plain request such as asking to show the Ponytail gain scoreboard.
- 5Read the benchmark bars. Compare each bar against the no-skill baseline and note the exact percentage labels.
- 6Use the repo commands for your code. For per-repo figures, run /ponytail-debt or /ponytail-audit inside your repository.
What an agent skill is and what ponytail-gain makes the agent do
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. The skill tells the agent what to do; it does not run on its own.
When invoked, ponytail-gain asks the agent to render a scoreboard. The figures are the published agentic benchmark of Ponytail 5: headless Claude Code (Opus 5.5, default effort) on 39 tasks, run 5 times each, against the same agent without the skill. The skill states that 18 of those tasks have hidden checks for correctness and safety.
The display is one-shot. The skill instructs the agent not to change mode, write flag files or persist anything.
- Bars show Ponytail as a share of the no-skill baseline for lines of code, output tokens, cost and time.
- Pass rates are shown for hidden correctness and safety checks, and for tests where the logic needs one.
- Each figure is the geometric mean of per-task medians, according to the source.
Install ponytail-gain from the ZIP or the Claude apps
Download the ZIP for this skill from this page and unzip it. The skill is a single file, SKILL.md. For a personal install across all your projects, place the unzipped skill folder inside ~/.claude/skills/. For a single project, place it inside that project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills
Trigger the scoreboard with a natural request
You do not need a special command. Ask in plain language for the Ponytail benchmark or impact summary, for example asking to show the Ponytail gain scoreboard. The agent should match the request to the skill's description, which mentions the scoreboard, less code, less cost, more speed and the agentic benchmark averages.
The skill's own output points to two other commands for work inside your repo. Those are /ponytail-debt, which shows shortcuts you deferred, and /ponytail-audit, which shows what is still cuttable.
- Show me the Ponytail gain scoreboard.
- What does the agentic benchmark say about the Ponytail skill?
- Give me the published Ponytail numbers for lines of code, cost and time.
Read the scoreboard correctly
The scoreboard shows benchmark averages, not your repository. The skill says it does not print per-repo savings, because there is no real baseline to subtract from in a live repo. The only per-repo figures come from /ponytail-debt, which the skill describes as a counted ledger.
Read the bars as a share of the no-skill baseline. A 47% lines-of-code bar means the benchmark average for Ponytail is 47% of the no-skill figure, which the skill labels as a 53% reduction.
Safety notes: scripts and network
According to the listing, ponytail-gain runs no scripts and uses no network access. It requires nothing beyond the agent itself. It does not change settings, write files or save anything.
The figures come from the project's own published benchmark and have not been independently checked in these files. Treat them as the project's reported results.
- Runs scripts: no
- Uses network: no
- Requires: nothing listed
- Licence: MIT
Tips for getting value from the skill
Use ponytail-gain before installing the wider Ponytail skill if you want to read the published summary first. It is also useful for seeing how a skill's impact can be reported, including the split between hidden checks and the logic-tests metric.
If you want numbers for your own code, do not rely on this scoreboard. Use the commands the skill points to, /ponytail-debt and /ponytail-audit, inside the repo you care about.
- Compare the Overall bars with the pass-rate lines before drawing conclusions.
- Remember the sample: 39 tasks, 5 runs each, one model configuration.
- Do not treat the averages as a forecast for your project.
What people use it for
- Reading the published Ponytail benchmark summary before deciding whether to install the wider skill.
- Showing a team how a skill's impact on code, cost and time is reported.
- Comparing the lines-of-code and cost bars against the pass-rate lines for hidden correctness and safety checks.
- Learning how benchmark averages differ from measurements in a live repository.
Frequently asked questions
How do I install the ponytail-gain skill?
Download the ZIP from this page and unzip it. Put the folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder for one project. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
Does ponytail-gain require anything else to run?
The listing states it requires nothing and runs no scripts and uses no network. It is a one-shot display of published figures.
Is ponytail-gain free and what is its licence?
The skill is licensed under MIT, as stated in the listing. The source does not mention any price.
Does the scoreboard show how much I save in my own project?
No. The skill says the figures are benchmark averages and it will not print per-repo savings. For real per-repo figures, use /ponytail-debt, which the skill describes as a counted ledger.
What are the alternatives if I want measurements for my own code?
The skill points to /ponytail-debt and /ponytail-audit for work inside a repo. The listing does not name other tools, so this guide does not recommend any.
Why does the scoreboard not match what I see in my project?
The figures come from a separate benchmark of 39 tasks, each run 5 times, with Claude Code on Opus 5.5 at default effort. Your tasks and setup will differ, so the averages are not a forecast for your repo.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use ponytail-gain if you want to read the published Ponytail benchmark summary quickly, or to see how a skill's impact is reported. Pick something else, or the repo commands /ponytail-debt and /ponytail-audit, if you need savings measured in your own code, since this skill shows only pre-published averages that have not been independently verified here.
Download the skill