release-notes
/release-notes
Pisze zwięzłe notatki do wydania z listy PR-ów, grupując zmiany tematycznie i cytując każdy PR dokładnie raz, z weryfikacją skryptem.
- Autor
- sejas ↗
- Repozytorium
- ai-code-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 08.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 08.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Pisze zwięzłe notatki do wydania z listy PR-ów, grupując zmiany tematycznie i cytując każdy PR dokładnie raz, z weryfikacją skryptem.
Oryginalny opis (EN)
Use when writing or polishing release notes for a version from a list of merged PRs (e.g. RELEASE-NOTES.txt, CHANGELOG). Produces concise, user-facing, thematically grouped notes where every line covers one coherent change and every PR is cited exactly once, with scripted verification.
Jak działa
- Workflow
- Pitfalls
Podgląd SKILL.md
---
name: release-notes
description: Use when writing or polishing release notes for a version from a list of merged PRs (e.g. RELEASE-NOTES.txt, CHANGELOG). Produces concise, user-facing, thematically grouped notes where every line covers one coherent change and every PR is cited exactly once, with scripted verification.
---
# Polished Release Notes
Turn a list of merged PRs into release notes where each line is concise, user-facing, and cites only truly related PRs. Never grab-bag unrelated fixes into one line.
## Workflow
### 1. Collect the PR set
- If a draft section already exists in the notes file, extract every `#NNNN` from it — that is the authoritative PR set.
- Otherwise list merges since the last release tag: `git log <last-tag>..HEAD --oneline | grep -oE '#[0-9]+'`. This also matches issue refs in commit subjects ("fixes #123") — sanity-check each number resolves to a merged PR via `gh pr view` before treating the set as authoritative.
- Fetch titles: `gh pr view <num> --json title,body`. Fetch bodies only for PRs whose title is ambiguous.
- Optional version column: `git tag --contains $(gh pr view <num> --json mergeCommit -q .mergeCommit.oid) | sort -V | head -1` gives the first tag containing each PR.
### 2. Draft (3 angles via subagents, or single-pass for small sets)
For >15 PRs, launch 3 parallel subagents, each drafting the full list from a different angle, then pick/synthesize:
- **Impact-first**: each line answers "what changed for me as a user?"
- **Minimalist**: shortest lines, split aggressively, multi-PR lines only for a feature + its direct follow-up
- **Grouped** (usually the winner): headline feature first, then cluster by product area; one bullet per coherent sub-theme
Each agent prompt MUST include: the full PR number→title list, the output format rules below, the style reference (3–5 lines quoted from the previous release section), and the instruction to end with a self-check line (`Check: N/N PRs used, no duplicates`). Agents may run read-only `gh pr view`; they must not edit files or post to GitHub.
For ≤15 PRs, draft the grouped angle directly yourself.
### 3. Line rules
- Format: `* Sentence #NNNN, #NNNN` — PR refs at end, comma-separated, match the file's existing conventions exactly (trailing period or not, prefix labels like `Studio Code:`, heading style).
- One coherent theme per line. Related PRs together (feature + follow-up fix, N PRs of one visual cleanup). Unrelated PRs never merged — no "and other improvements" grab-bags.
- Plain user-facing language. No implementation jargon (IPC, daemon socket, mu-plugins) unless the audience is developers and there is no user-visible phrasing.
- Order: headline feature → other features → improvements → fixes → maintenance/deps.
- Keep product names and established terms exactly as the project writes them.
### 4. Verify (mandatory before applying)
Every PR appears exactly once — check by script, never by eye. Usage: `python3 check.py <notes-section.txt> <expected-prs.txt>`, where the expected file holds bare PR numbers (no `#`), whitespace-separated — e.g. built with `grep -oE '#[0-9]+' old-section.txt | tr -d '#' | sort -u > expected-prs.txt`:
```python
import re, sys
expected = {n.lstrip('#') for n in open(sys.argv[2]).read().split()}
nums = re.findall(r'#(\d+)', open(sys.argv[1]).read())
assert len(nums) == len(set(nums)), f"duplicates: {sorted(n for n in nums if nums.count(n)>1)}"
assert set(nums) == expected, f"missing: {sorted(expected-set(nums))} extra: {sorted(set(nums)-expected)}"
print(f"OK {len(nums)}/{len(expected)}")
```
Do not trust a subagent's self-reported check — re-run the script on the final text yourself.
### 5. Apply
- Replace only the target version's section in the notes file.
- Re-run the verification script against the file section after editing.
- Do not commit unless the user asks.
## Pitfalls
- A reworded line silently dropping one PR of a group — the script catches this; run it after every edit.
- Copying a subagent draft that added trailing periods or headings the file doesn't use — match the file, not the agent.
- Bundling a "polish" mega-line: if a line needs "and other details" to close, split it.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · sejas/ai-code-skills | 100 | MIT |
| mono/SkiaSharp | 100 | MIT |
| niekcandaele/claude-helpers | 100 | MIT |
| kenryu42/cc-safety-net | 100 | MIT |
| zeitlos/lucity | 100 | AGPL-3.0 |
| openshift-pipelines/pipelines-as-code | 100 | Apache-2.0 |
| paperwingio/paperwing | 100 | Apache-2.0 |
| jelvodev/yaap | 100 | MIT |
| yearzen1/LocalSkillManager | 100 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 075c9da50b2d (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100release-notes-writer
Przekształca scalone pull requesty, logi commitów i wpisy changelogu w czytelne notatki wydania dla użytkowników, pogrupowane na funkcje, ulepszenia, poprawki i zmiany łamiące zgodność. - 100v1-changelog
Pisze notatki wydania ze scalonych pull requestów, gdy użytkownik prosi o changelog albo podsumowanie wydania. - 100create-release-notes
Generuje notatki wydania dla użytkowników: dokumentuje release, streszcza zmiany i przygotowuje changelog między wersjami. - 100release-notes-generator
Generuje czytelne notatki wydania z PR-ów, commitów lub changelogów między wersjami, grupując zmiany i wyróżniając zmiany łamiące. - 100release-notes-writer
Tworzy gotowe do wysłania notatki wydania ze specyfikacji i zgłoszeń, zarówno dla użytkowników, jak i wewnętrzne dla inżynierów. - 100release-notes-changelog-communication
Pisze przyjazne dla użytkowników notatki o wydaniu, opisujące nowe funkcje, poprawki błędów i zmiany łamiące zgodność.