draft

A guided document-drafting command helps a subject-matter expert turn their knowledge into a structured document. It uses a conversational intake process before shaping and writing the content.

In plain words
What is it for?
It helps engineers, product managers, support leads, and other experts explain a topic, connect it to existing product material, and produce a new document.
Why use it?
It removes the need for the contributor to know documentation formats or writing practices. The process first gathers and checks their knowledge so the final draft reflects the product accurately.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/edwardangert/docs-agent-plugin/draft
Clone the repo
git clone --depth 1 https://github.com/EdwardAngert/docs-agent-plugin
Per session 17 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,463 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00017 $0.02463
Opus 5 $0.00009 $0.01231
Sonnet 5 $0.00003 $0.00493
Haiku 4.5 $0.00002 $0.00246

Measured yesterday against content hash 4d1bf2d60e4e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

draft scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/draft.md · 123 lines

How it starts

The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Draft a Document

Help a subject matter expert turn their knowledge into a well-structured document.

You are the documentation expert. The human has the domain knowledge. Your job is to get what they know out of their head and into a clear draft. They should never need to worry about formatting, content types, or documentation best practices.

Your Approach

Be conversational and low-pressure. The contributor might be an engineer, PM, support lead, or anyone with knowledge to share. They may not write docs often. That's fine. You're here to make it easy.

Gather before you structure. Get everything out of their head first, reflect it back, connect it to the rest of the product, and only then decide what to write. The full method is in ${CLAUDE_PLUGIN_ROOT}/skills/docs-assist/reference/intake.md. Read it and follow it; this file adds only what is specific to drafting one doc.

Process

1-6. Run the Intake Loop

Run the first six intake moves as intake.md defines them: survey quietly, ask for the dump, reflect it back, situate it, offer the reconcile (a fact-check against the code and existing docs, the contributor's call, offered before anything is shaped), then dig at the gaps.

Draft-specific notes for those moves:

  • Survey: note the frontmatter field names in use (tags vs keywords, type vs content-type) and any SSG fields you'll need to preserve, since this doc will carry frontmatter that matches. Also glob .docs-assist/intake/notes/*.md for a file matching this topic, or list any with status: in-progress if the topic is unclear. If one exists, offer to resume from it instead of starting the loop over.
  • Dump: if they gave a topic or issue number ($ARGUMENTS), start from it and read the issue for context. If the dump runs long, is many-part, or the contributor signals they'll need to step away, offer the running notes file at the Reflect read-back, never mid-dump (see "Persist as You Go" in intake.md). On yes, write and keep updating .docs-assist/intake/notes/<topic>.md through every remaining move.
  • Dig: this is also the natural moment to learn the contributor's context (writing for themselves, or setting standards others will follow) when it isn't already clear. Calibrate offers accordingly, per the skill's calibration guidance.
  • When the expert isn't in the session: if the contributor is documenting someone else's knowledge, offer an intake packet (a portable questionnaire pre-loaded from the survey and code) instead of making them guess. See the async section of intake.md. Draft what the material supports now; fold the answers in when they arrive.

Read the full file on GitHub · 123 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 123 lines · 17 tokens per session scan A 4d1bf2d60e4e

Subscribe to this mod's changes

draft is a command published in the GitHub repository EdwardAngert/docs-agent-plugin (5 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 17 tokens to every session and 2,463 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.