Skip to content
HUIAHUB

anthropics/skills/skills/doc-coauthoring

Document co-authoring

Guides users through a structured process for writing documentation, proposals, technical specs, decision documents, and similar substantial content.

Agent skillAgent skills
View the skill on GitHub (opens in a new tab)

What Document co-authoring does

The skill guides Claude through three stages. It starts by gathering the document type, audience, intended effect, format constraints, background, technical details, and other relevant context. Claude asks follow-up questions about gaps and can read supplied files or retrieve shared documents through available integrations.

Next, Claude builds the document section by section. It asks focused questions, proposes possible content, lets the user choose what to keep or combine, drafts the section, and applies targeted edits until the user is satisfied. It recommends starting with the section that has the most unknowns and leaving summaries until later.

The final stage checks how the document reads without the conversation context. When sub-agents are available, Claude gives a fresh instance the document and predicted reader questions, then checks for misunderstandings, assumptions, ambiguity, and contradictions. Without sub-agents, the skill gives the user instructions for running the same checks in a fresh Claude conversation. Problems return to the refinement stage before the user's final review.

Install

/plugin install example-skills@anthropic-agent-skills

Read the security check below before you run it. The repository's own instructions win if they differ.

Security check

What the skill's files do, from reading them. A flag that nobody has checked yet says "Not checked", never "No".

Runs scripts
No
Uses the network
Yes
Installs hooks
No
Asks for secrets
No

The scanned skill contains no scripts, installs no hooks, and does not require secrets. It can use network access because its instructions allow fetching linked shared documents through available integrations and direct users to Claude.ai for manual reader testing. Before installing, read the SKILL.md and check which connected services your agent can access. Review any requested document or messaging integration before allowing it to retrieve project context.

Scan details: SKILL.md instructions include web requests or URLs to fetch. Checked at commit 8a1541c4a3ff.

Use cases

  • Draft technical specifications and design documents.
  • Develop proposals, PRDs, RFCs, and decision documents.
  • Refine a substantial document section by section.
  • Check a finished draft from a fresh reader's perspective.

Limitations

  • Shared documents require an appropriate integration or pasted content.
  • Reader testing is manual when sub-agents are unavailable.
  • The process depends on the user supplying enough project and audience context.
  • The user remains responsible for checking facts, links, and technical details.

Sources