uplika/uplika-plugin / skills/yt-package
yt-package
VerifiedWrite and lint a YouTube title and thumbnail as one pairing (truncation, repeats, vagueness), then apply it. For titles, thumbnail text, low CTR.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.9 KB |
| scripts/title.py | script | 5.5 KB |
| tests/test_title_ko.py | script | 2.2 KB |
scripts/title.py
View on GitHub#!/usr/bin/env python3
"""title.py - lint a YouTube title and thumbnail pairing before you publish it.
python3 title.py --title "..." --thumb "AI RAN IT"
python3 title.py titles.txt # one per line, ranked
python3 title.py --title "..." --json
The pairing is the unit, not the title. A title that repeats the thumbnail text wastes half the
click surface, and that is the single most common mistake this checks for.
Length: YouTube truncates around 60 characters on desktop search and around 40 on a mobile home
feed. Both limits are reported because they are different failures - a desktop truncation loses the
tail, a mobile one can lose the subject.
"""
# Adapted from Jakeschincariol/youtube-agent-skill (MIT) for the Uplika plugin. English behaviour is
# the original. Korean (2026-10-10) is heuristic, NOT calibrated: words are space-separated eojeol
# with particles attached, the capitals check skips Hangul words (Hangul has no capitals, and an
# acronym with a particle such as GPT로 is not shouting), and the cut points stay the ones measured
# on Latin text. The module docstring above is the CLI's usage text, so it stays the original.
import json, re, sys, os
DESKTOP, MOBILE, HARD = 60, 40, 100
VAGUE = {"amazing","incredible","insane","crazy","huge","massive","ultimate","best","powerful",
"secret","revolutionary","mindblowing","epic","perfect","complete","everything"}
STOP = {"the","a","an","of","for","to","in","on","and","or","is","are","with","your","you","my","i",
"this","that","it","how","what","why"}
HANGUL = re.compile(r"[가-힣]")
# Hype and intensifier words, matched at the start of a word because endings attach (미쳤다, 엄청난)
KO_VAGUE = re.compile(r"^(?:진짜|완전|대박|미쳤|레전드|엄청)")
# The original tokenizer was [a-z0-9']+. Adding the Hangul range changes nothing for text without Hangul.
def words(t): return re.findall(r"[가-힣a-z0-9']+", t.lower())
def ko_shared(a_words, b_words):
"""Korean words two texts share, allowing a particle on one side: 자동화 and 자동화하는, 쇼츠 and 쇼츠를."""
return {min(a, b, key=len) for a in a_words for b in b_words
if HANGUL.search(a) and HANGUL.search(b) and min(len(a), len(b)) >= 2
and (a.startswith(b) or b.startswith(a))}
def check(title, thumb=None):
t = title.strip()
ko = bool(HANGUL.search(t) or (thumb and HANGUL.search(thumb)))
n = len(t)
issues, good = [], []
if n > HARD: issues.append(("length", f"{n} characters - YouTube's hard limit is {HARD}"))
elif n > DESKTOP: issues.append(("length", f"{n} characters - desktop search cuts near {DESKTOP}"))
else: good.append(f"{n} characters, inside the {DESKTOP}-character desktop cut")
if n > MOBILE:
head = t[:MOBILE].rsplit(" ", 1)[0]
issues.append(("mobile", f'a mobile feed shows about "{head}..." - check the subject survives'))
caps = [w for w in t.split() if len(w) > 2 and w.isupper() and not HANGUL.search(w)]
if len(caps) > 2: issues.append(("shouting", f"{len(caps)} all-caps words - two is the ceiling before it reads as spam"))
elif caps: good.append(f"{len(caps)} all-caps word for emphasis")
v = [w for w in words(t) if w in VAGUE]
if ko: v += [w for w in words(t) if KO_VAGUE.match(w)]
if v: issues.append(("vague", f"{', '.join(sorted(set(v)))} - swap for a number, a name or a date"))
nums = re.findall(r"\d[\d,.]*%?", t)
if nums: good.append(f"carries a concrete figure ({', '.join(nums[:3])})")
else: issues.append(("no-number", "no number, date or name - the most reliable single fix"))
if t.endswith("?"): good.append("open question in the title")
front = [w for w in words(t)[:3] if w not in STOP]
if not front: issues.append(("front-load", "the first three words are all filler - move the subject forward"))
if thumb:
tw, hw = set(words(t)) - STOP, set(words(thumb)) - STOP
shared = tw & hw
if ko: shared |= ko_shared(tw, hw)
if shared:
issues.append(("duplicate", f"thumbnail repeats the title on {', '.join(sorted(shared))} - "
"the thumbnail should say what the title does not"))
else:
good.append("thumbnail and title carry different words")
if len(words(thumb)) > 4:
issues.append(("thumb-length", f"{len(words(thumb))} words on the thumbnail - three is the ceiling at feed size"))
score = max(0, min(100, 100 - 14 * len(issues) + 4 * len(good)))
r = {"title": t, "chars": n, "score": score, "issues": issues, "good": good}
if ko: r["calibration"] = "none"
return r
def show(r):
print(f'\n "{r["title"]}"')
print(f" {r['chars']} chars score {r['score']}/100")
for k, m in r["issues"]: print(f" x {k:<12} {m}")
for m in r["good"]: print(f" ok {m}")
if r.get("calibration") == "none":
print(" Korean: heuristic, not calibrated (the cut points were measured on Latin text, and Hangul is wider)")
def main():
a = sys.argv[1:]
as_json = "--json" in a; a = [x for x in a if x != "--json"]
thumb = a[a.index("--thumb") + 1] if "--thumb" in a else None
if "--title" in a:
rows = [check(a[a.index("--title") + 1], thumb)]
elif a and os.path.exists(a[0]):
rows = [check(l, thumb) for l in open(a[0]).read().splitlines() if l.strip()]
else:
print(__doc__); sys.exit(1)
rows.sort(key=lambda r: -r["score"])
if as_json: print(json.dumps(rows, indent=1)); return
for r in rows: show(r)
print()
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |