doc-coauthoring
Drafts a long document with you in passes — outline, then sections — instead of dumping four thousand words you have to cut apart.
npx skills add anthropics/skills --skill doc-coauthoringTwo skills that change how a document gets written: one builds a long document with you in passes, the other keeps recurring updates in one voice.
Neither one produces a finished document in a single request. doc-coauthoring expects an outline you have agreed on before it writes a section; internal-comms expects your voice and your usual sections to be described first. Give them that and the draft needs editing instead of rewriting.
Drafts a long document with you in passes — outline, then sections — instead of dumping four thousand words you have to cut apart.
npx skills add anthropics/skills --skill doc-coauthoringWrites internal updates, status reports and announcements in one consistent voice, so your team does not get a different tone every month.
npx skills add anthropics/skills --skill internal-commsNo skill in this category has a full write-up yet.
doc-coauthoring is for a document that does not exist yet. It works in passes: you agree on a structure, it writes a section, you correct that section, it continues. That suits specs, reports, proposals and anything long enough that a once-through draft would drift halfway. internal-comms is for a document that exists every week — status updates, launch notes, incident summaries. Its job is consistency: the same sections, the same tone, the same level of detail, so a reader can compare this month with last month without re-learning the format. Pick doc-coauthoring when the hard part is thinking, and internal-comms when the hard part is repeating yourself in the same shape. If your team writes both kinds of document, install both — they do not overlap.
Read the draft against your own outline first: every heading covered, and nothing invented in a section you did not ask for. Then check what a model cannot know — numbers, dates, names of people and products, links — against the source. For internal-comms, put last month's update next to this one: if the sections, the length or the tone do not line up, the voice description inside the skill is too loose and needs tightening. Neither skill announces that it guessed; only a comparison against the source catches it.
Only if you write two different kinds of document. doc-coauthoring helps you build a long document that does not exist yet; internal-comms repeats a document you write regularly. A spec or a report is the first case. A weekly status update is the second.
It is built for passes, not one shot. Asking for the finished document usually produces a long draft whose structure was guessed rather than agreed, and that draft needs more editing than the pass-by-pass route. Give it an outline you actually want first.
The same place as every other skill on this site: run the install command on the row, or copy the folder to ~/.claude/skills/name/SKILL.md, where name is the skill's folder name. The path has to end at the SKILL.md file — an extra folder on top of it is the most common reason a skill never appears.
The folder format is the one Claude Code reads, and Codex, Cursor and Gemini CLI read the same format, which is why each row carries one command instead of four. Where a client behaves differently, its README says so — check it before installing.