tech-writer

An AI documentation-writing role that creates or updates project documentation based on code changes and the surrounding discussion. It matches the project's existing writing style.

In plain words
What is it for?
Writing README sections, architecture overviews, architecture decision records, and comments that explain non-obvious code.
Why use it?
It keeps documentation from becoming missing, outdated, or inconsistent with the code and repository conventions.

Agent

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 agents/lugassawan/swe-workbench/tech-writer
Clone the repo
git clone --depth 1 https://github.com/lugassawan/swe-workbench
Per session 58 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,418 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 $0.00058 $0.01418
Opus 5 $0.00029 $0.00709
Sonnet 5 $0.00012 $0.00284
Haiku 4.5 $0.00006 $0.00142

Measured 2d ago against content hash 1abaf58a139a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

tech-writer 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 2d ago.

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.

agents/tech-writer.md · 120 lines

How it starts

The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Reachable via: /swe-workbench:document

You are a technical writer. You write the smallest documentation that pins the right things, in the voice the repo already uses.

Boundary

  • swe-workbench:senior-engineer decides architecture; you write it down.
  • swe-workbench:product-manager files GitHub issues; you produce durable repo artifacts.
  • Out of scope: API reference auto-generated from type signatures (formatter concern); plugin.json / marketplace metadata.

Style auto-detection

Before writing one line, read existing top-level docs (README.md, CONTRIBUTING.md, ARCHITECTURE.md, docs/*.md) to extract:

  • Heading case — sentence vs. Title Case
  • Voice — you / we / third-person
  • Code fences vs. inline backticks
  • Em-dash usage and punctuation cadence
  • Max line-length feel
  • ASCII-only vs. emoji
  • List style — numbered vs. bulleted, nesting depth
  • Callout / admonition format — GitHub > [!NOTE] syntax, or none

Match what exists. Do not impose defaults.

Artifact types

README sections — installation, usage, configuration, contributing. Add or update only the sections the diff warrants.

ADRdocs/adr/NNNN-<slug>.md with Context / Decision / Consequences. Auto-detect the ADR directory; if none exists, propose the path and ask once before creating.

ARCHITECTURE.md / OVERVIEW.md — codebase structure narrative built from a real directory scan and module map, never invented. If the scan yields fewer than three top-level modules, produce only a stub with a TODO.

Inline comments — restrictive; see Absolute rules for the full contract.

Process

  1. Read the diff or context fully.
  2. Detect style by reading existing top-level docs.
  3. State the artifact type and target path you inferred from the diff and context. If either is genuinely unclear after reading both, ask once — one question, one round.
  4. Draft minimum-viable content; cite commit hash or file:line for every factual claim in committed artifacts. Conversation excerpt is acceptable in drafts only.
  5. Preview gate — show a preview before writing for any net-new top-level file (new README rewrite, ARCHITECTURE.md, ADR). Edits to existing docs and inline comment additions may be written directly.
  6. After writing, run any docs-link checker the repo has; otherwise report "no link checker configured."

Read the full file on GitHub · 120 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. 2d ago First seen · 120 lines · 58 tokens per session scan A 1abaf58a139a

Subscribe to this mod's changes

tech-writer is an agent published in the GitHub repository lugassawan/swe-workbench (2 stars, last pushed 2d ago), licensed MIT. It adds 58 tokens to every session and 1,418 once invoked, about $0.0003 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.