Revanced ponytail gain
Skill ChloeVPin/revanced-ponytail/skills/revanced-ponytail-gain
Show measured impact from the upstream agentic benchmark only: less code, less cost, more speed, with limits. One-shot display. Not a per-repo savings number. Trigger: /revanced-ponytail-gain, "ponytail gain", "what does ponytail save", "show impact scoreboard".From its SKILL.md
npx -y skills add ChloeVPin/revanced-ponytail --skill revanced-ponytail-gainAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 21 days oldThe repository was created 21 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.4 KB, 571 tokens by cl100k_base, as published. Nobody here has run it
revanced-ponytail-gain
Display this card when invoked. One-shot. Do not change mode. Do not invent per-repo savings.
Scoreboard (agentic, not single-shot)
Source: DietrichGebert/ponytail agentic benchmark, 2026-06-18.
Setup: Claude Code sessions, Haiku 4.5, real FastAPI + React template, n=4, LOC from git diff added lines.
12 feature tasks vs no-skill baseline (means):
| Arm | LOC | tokens | cost | time |
|---|---|---|---|---|
| ponytail | -54% | -22% | -20% | -27% |
| yagni-oneliner prompt | -33% | -14% | -21% | -30% |
| caveman (terse prose) | -20% | +7% | +3% | +2% |
Reading:
- Large cuts appear where over-build traps exist (for example custom date/color pickers vs native inputs).
- Near zero where code is already minimal.
- Peak per-task reductions can reach ~90%+ on over-build traps. That is not the mean.
Safety suite (6 surgical tasks, deterministic checks):
| Arm | Safe rate on suite |
|---|---|
| ponytail | 20/20 |
| baseline / caveman | 20/20 |
| yagni-oneliner | 19/20 |
This is a floor, not a security proof. One model. Small suite.
Honesty boundary
- These are published agentic medians, not this repository.
- Never print "you saved X lines here." The unbuilt version was never written.
- Do not cite the old single-shot 80-94% figure as the headline. That baseline was partly conversational padding.
- For local deferred work, point to
revanced-ponytail-debt. - For local cut candidates, point to
revanced-ponytail-audit.
ASCII card (optional render)
revanced-ponytail gain agentic · 12 feature tasks · Haiku 4.5 · n=4
LOC vs baseline ponytail mean -54% (peak where over-build exists)
tokens / cost / time -22% / -20% / -27%
safety suite 20/20 known guards kept (floor, not proof)
This repo: revanced-ponytail-debt | revanced-ponytail-audit
Boundaries
One-shot. Edits nothing. Changes no mode.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.