Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate user-facing release notes from tickets, PRDs, or changelogs. Creates clear, engaging summaries organized by category (new features, improvements, fixes). Use when writing release notes, creating changelogs, announcing product updates, or summarizing what shipped.
.claude/skills/phuryn-release-notes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-18 | ✓→✗ | ▼ Worse | -9% | 0% |
| case-20 | ✓→✗ | ▼ Worse | -9% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 3% | 0% |
Transform technical tickets, PRDs, or internal changelogs into polished, user-facing release notes.
You are writing release notes for $ARGUMENTS.
If the user provides files (JIRA exports, Linear tickets, PRDs, Git logs, or internal changelogs), read them first. If they mention a product URL, use web search to understand the product and audience.
Example transformations:
# Product Name] — Version / Date]
## New Features
## Improvements
## Bug Fixes
## Breaking Changes (if any)
Save as a markdown document. If the user wants HTML or another format, convert accordingly.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 9,264 | 6,685 | -28% | 1 | 1 | 0% | 1,642 | 1,515 | -8% | 0 | 0 | — |
case-01 | fail→fail | 7,949 | 5,455 | -31% | 1 | 1 | 0% | 1,503 | 1,536 | +2% | 0 | 0 | — |
case-03 | fail→fail | 8,884 | 4,337 | -51% | 1 | 1 | 0% | 1,565 | 1,338 | -15% | 0 | 0 | — |
case-04 | pass→pass | 10,733 | 5,489 | -49% | 1 | 1 | 0% | 1,461 | 1,502 | +3% | 0 | 0 | — |
case-05 | pass→pass | 8,489 | 9,164 | +8% | 1 | 1 | 0% | 1,346 | 1,600 | +19% | 0 | 0 | — |
case-06 | fail→fail | 5,095 | 4,350 | -15% | 1 | 1 | 0% | 764 | 1,284 | +68% | 0 | 0 | — |
case-07 | pass→pass | 7,002 | 4,122 | -41% | 1 | 1 | 0% | 997 | 1,186 | +19% | 0 | 0 | — |
case-08 | fail→pass | 13,325 | 5,086 | -62% | 1 | 1 | 0% | 2,910 | 1,473 | -49% | 0 | 0 | — |
case-09 | fail→fail | 8,175 | 7,501 | -8% | 1 | 1 | 0% | 1,497 | 1,629 | +9% | 0 | 0 | — |
case-10 | pass→pass | 10,465 | 5,431 | -48% | 1 | 1 | 0% | 1,607 | 1,503 | -6% | 0 | 0 | — |
case-11 | pass→pass | 9,032 | 4,025 | -55% | 1 | 1 | 0% | 1,733 | 1,148 | -34% | 0 | 0 | — |
case-12 | pass→pass | 8,166 | 6,727 | -18% | 1 | 1 | 0% | 1,400 | 1,380 | -1% | 0 | 0 | — |
case-13 | pass→pass | 7,009 | 4,793 | -32% | 1 | 1 | 0% | 1,246 | 1,424 | +14% | 0 | 0 | — |
case-14 | pass→pass | 6,814 | 5,582 | -18% | 1 | 1 | 0% | 1,357 | 1,467 | +8% | 0 | 0 | — |
case-15 | fail→pass | 7,298 | 4,028 | -45% | 1 | 1 | 0% | 1,215 | 1,199 | -1% | 0 | 0 | — |
case-16 | pass→pass | 8,141 | 5,019 | -38% | 1 | 1 | 0% | 1,577 | 1,487 | -6% | 0 | 0 | — |
case-17 | pass→pass | 5,094 | 3,902 | -23% | 1 | 1 | 0% | 922 | 1,202 | +30% | 0 | 0 | — |
case-18 | pass→fail | 8,428 | 7,175 | -15% | 1 | 1 | 0% | 1,517 | 1,380 | -9% | 0 | 0 | — |
case-19 | pass→pass | 23,757 | 16,423 | -31% | 1 | 1 | 0% | 4,163 | 3,382 | -19% | 0 | 0 | — |
case-20 | pass→fail | 23,249 | 18,187 | -22% | 1 | 1 | 0% | 4,136 | 3,771 | -9% | 0 | 0 | — |
case-21 | pass→pass | 18,623 | 13,660 | -27% | 1 | 1 | 0% | 3,332 | 3,839 | +15% | 0 | 0 | — |
case-22 | pass→pass | 4,376 | 4,498 | +3% | 1 | 1 | 0% | 677 | 1,191 | +76% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of 0 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.