writing-docs
Skill0 starsChanged 6 days agoMITView on GitHub
Write CHANGELOG.md and README.md entries for this extension — facts only, no rationale, no prose. Use before adding a line to [Unreleased], before editing README.md, when stamping a release section, or when the user says "запиши в чейнджлог", "поправь ридми", "напиши changelog", "update the README".
22 other projects use a version of this.
Adopted by
Most recently taken up by
- johannschopplich/skillstheir own version, first seen Nov 8, 2026
- The-Vibe-Company/quivrtheir own version, first seen Nov 8, 2026
- Wade-DevCode/awesome-coding-skills-cntheir own version, first seen Nov 8, 2026
- gabrielbauman/skillstheir own version, first seen Nov 1, 2026
- stormlightlabs/miretheir own version, first seen Nov 1, 2026
- NVIDIA/Megatron-Energontheir own version, first seen Nov 1, 2026
Used with
Projects with it also use pr-description, implement and 5 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
MicrosoftSkillpr-description
Draft Nacelle PR titles, pull request descriptions, commit titles, commit descriptions, and squash commit messages. Use when asked for PR or commit wording, a branch summary, or a PR description that can be used in a commit. Omit validation sections from commit bodies and remove filler.
Also in 3 projects that use this
Hugging FaceSkillimplement
Make tests pass. Invoke after /write-tests produces failing tests.
Also in 5 projects that use this
Google CloudSkilltriage
Gather context on a change or bug and plan the implementation.
Also in 3 projects that use this
GoogleSkillprototype
Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.
Also in 3 projects that use this
GoogleSkillgrilling
Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.
Also in 3 projects that use this
MicrosoftSkillhandoff
Compact the current conversation into a handoff document for another agent to pick up.
Also in 3 projects that use this
- +1 more from google
Other versions22 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 22
- remotion-dev/remotion★ 72ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- medusajs/medusa★ 37ksame name
- Nagi-ovo/gemini-voyager★ 21ksame name
- nukeop/nuclear★ 19ksame name
- ngrx/platform★ 8.3ksame name
- cuthbertLab/music21★ 2.6ksame name
- c15t/c15t★ 2ksame name
- NVIDIA/Megatron-Energon★ 388same name
12 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 21 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 27
Keep a Changelog, Conventional Commits, pnpm, Svelte and Astro, on 2026-08-24:
- Very specific: line 86
Keep a Changelog's six: Added, Changed, Deprecated, Removed, Fixed, Security.
“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Shared execution contract
Use the current client's real tools and the repository AGENTS.md. Historical tool names are examples, not callable API guarantees. Git/publication, messaging another chat and destructive actions require applicable user authorization; never include unrelated unfinished work. Model/effort and delegation follow the actual session. Verify the resulting artifact before reporting success.
Writing the CHANGELOG and the README
Both files are read by users of the extension, not by its maintainers. They carry what the software does and what changed in it. Nothing else — no reasoning, no justification, no story about how the change came about.
The rule is one line: a changelog entry states the fact; the reason lives in the commit message and in the code comment. This repo already comments its code densely with why — that is where reasoning belongs, and it is why none of it is needed here.
The measured norm
Sampled from eight VS Code extensions (GitLens, Prettier, ESLint, Docker, Error Lens, Live Server, Vim, Python), ten applications and CLI tools (ripgrep, bat, fd, curl, jq, Node, Zed, Ghostty, Obsidian, Bun) and the stated rules of Keep a Changelog, Conventional Commits, pnpm, Svelte and Astro, on 2026-08-24:
| Median bullet | Max | Entries carrying a reason | |
| Prettier, ESLint, Docker, Error Lens, Vim, Python | 7–11 words | 20–48 | 0–2.5% |
| curl, Node, ripgrep, fd, jq | 8–12 words | — | ~0% |
| Zed, bat — the explanatory end of the range | 16 words | — | up to 25% |
| GitLens (marketing-style outlier) | 14 | 122 | 20% |
| This repo, before this skill | 21 | 142 | 22% |
pnpm states the split outright: "Write the description for pnpm users and keep
it concise — it becomes a release note. Implementation rationale belongs in the
commit message, not the changeset." Svelte enforces it structurally: the
changelog gets fix: description, while "Why make this breaking change" is a
field in the PR template.
CHANGELOG entries
Form. One change, one bullet, one line where possible. Past tense, matching the section it sits under. No trailing period is fine; be consistent. Target ≤ 15 words, ceiling 20. A bullet may run to about 30 when every word of it is a named fact — a list of the values a setting takes, or a row of measurements. Past that, split it in two.
Content. What changed, where it is visible. Name the tab, the setting, the
command, the status-bar field. A number is a fact and always welcome —
281 of 361 were hidden is exactly right.
Banned in an entry:
- Any clause answering why we did it:
because,so that,which is how,the reason,rather than …,otherwise. - Any claim about what the change is worth: "worth", "cannot afford", "half delivered", "silently", "rots".
- Any account of the bug's history, of what a reader would have thought, or of what shipped before.
- Em-dash asides that turn one fact into two clauses.
- Design narration — how the code is built. That belongs in
AGENTS.md.
Three exceptions, and only these — matching where the sampled projects do carry a sentence of explanation:
- A breaking change → say what the user must do, not why it broke.
- A default that flipped → say the new default and how to restore the old one.
- A security fix → say what was exposed.
Even then it is one sentence about the user's next action, never about our reasoning.
Before and after, from this file
| Wrong | Right |
| The Launch tab's Extra arguments field was drawn by the browser rather than from the theme — a black box on a light background, which shipped in 0.42.0 and stood in the listing screenshot. | Extra arguments field on the Launch tab is themed; it was drawn by the browser. |
| Each of the three figures now names the measurement it comes from, under the finding it decides: … three separate measurements, no two of them commensurable. A shared footer naming two documents left the reader unable to tell which number came from which. | Each figure in the Launch verdict names its benchmark, size and method. |
| A session opens in the first folder of the workspace instead of wherever the last editor happened to be. The button never said which directory it wanted, so in a multi-root window VS Code answered from the editor history — and a Claude session's directory is its identity. | Open Claude Code starts the session in the workspace's first folder, not in the last editor's folder. |
Sections
Keep a Changelog's six: Added, Changed, Deprecated, Removed, Fixed,
Security. Documentation is ours, for README and screenshot work. One section
per kind per release — never two ### Added blocks in one version.
Not in the changelog at all: refactors, tests, comments, internal rules, dependency bumps nobody can observe, anything invisible from the UI.
README
It is the marketplace listing. It answers what this is, what it shows, how to run it, what it reads — in that order, with facts.
Skeleton (the shape every reference-style tool README in the sample uses): definitional sentence → install → what it does, enumerated → configuration reference → data sources → known limits → contributing → licence.
The opening sentence is a definition, inside ten words: "X is a Y that does
Z." Microsoft's publishing docs mandate nothing about tone; the only hard rules
are that images resolve over https and are not SVG.
Banned in the README:
- Argument that the extension should exist — "nothing else adds this up", "almost none of it", comparisons meant to convince.
- Narrative about the problem before the description of the tool.
- Explanations of internal design: which module renders what, why two surfaces share one function, what the page cannot do. Users do not maintain this code.
- Adjectives doing work a number could do. Give the number.
Welcome in the README: tables of tabs, settings, commands and fields; placeholder syntax; screenshots; exact file paths that are read; the list of what is never read; version and platform requirements.
Check before you commit
# Bullets longer than 25 words in the section you just wrote
python3 - <<'PY'
import re
cur = open('CHANGELOG.md').read().split('## [')[1]
for b in re.findall(r'^- .+?(?=\n- |\n### |\n## |\Z)', cur, re.S | re.M):
n = len(b.split())
if n > 25: print(f'{n:3} words: {" ".join(b.split())[:110]}…')
PY
# Reason-words in the section you are writing, and in the whole README
REASONS='\b(because|so that|which is how|the reason|rather than|otherwise|worth|cannot afford)\b'
sed -n '/## \[Unreleased\]/,/^## \[0/p' CHANGELOG.md | rg -n --no-heading -i "$REASONS"
rg -n --no-heading -i "$REASONS" README.mdBoth should print nothing outside the three exceptions above. If a line survives the grep, it either states a user action (keep it) or explains us (cut it).
Released sections are left as they are: they are the release notes users already
received. The whole file was normalised to this standard once, on 2026-08-24,
facts unchanged — that was a one-off, not a licence to edit published notes
again. The check is scoped to [Unreleased] for that reason.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

