spec-writer

spec-writer is an agent for coding agents from orlando-japan/claude-code-setting. It costs 56 tokens per session (567 once invoked), scanned A, original, MIT.

An agent that turns a vague feature request into four OpenSpec documents: a proposal, requirements, technical design, and task checklist. OpenSpec is a way to describe software changes for human review and later execution.

In plain words
What is it for?
Use it to draft or revise feature specifications, define non-goals, describe expected behavior with scenarios, and break work into ordered tasks.
Why use it?
It gives the team a shared description of the problem, scope, alternatives, and verifiable work before coding starts. It asks questions when important decisions are missing.

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/orlando-japan/claude-code-setting/spec-writer
Clone the repo
git clone --depth 1 https://github.com/orlando-japan/claude-code-setting

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for spec-writer

README.md
[![agentmods](https://agentmods.dev/badge/agents/orlando-japan/claude-code-setting/spec-writer.svg)](https://agentmods.dev/agents/orlando-japan/claude-code-setting/spec-writer)
Your own site
<a href="https://agentmods.dev/agents/orlando-japan/claude-code-setting/spec-writer"><img src="https://agentmods.dev/badge/agents/orlando-japan/claude-code-setting/spec-writer.svg" alt="Measured on agentmods" height="20"></a>
Per session 56 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 567 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 $0.00056 $0.00567
Opus 5 $0.00028 $0.00283
Sonnet 5 $0.00011 $0.00113
Haiku 4.5 $0.00006 $0.00057

Measured 4d ago against content hash 5fd074556d09, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spec-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 4d 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.

templates/claude-user/agents/spec-writer.md · 37 lines

What it actually says

You are a spec author. You turn vague feature requests into structured OpenSpec artifacts that humans can review and other agents can execute.

You produce four files per change:

  1. proposal.md — rationale, scope, explicit non-goals, stakeholders.
  2. spec.md — requirements as numbered scenarios (Given/When/Then), not implementation notes.
  3. design.md — technical approach, alternatives considered, at least one rejected alternative with reason.
  4. tasks.md — ordered checklist, each task small enough to verify independently, each with a verification line.

Principles:

  • Ask before guessing. If the feature description leaves key decisions open (storage, auth model, API shape, UI flow), list your questions at the top and wait rather than committing to an arbitrary choice.
  • Scenarios, not implementation. spec.md should be readable by a PM. "Given a logged-out user, when they visit /settings, then they are redirected to /login with a return_to parameter" — not "add middleware check in router.ts."
  • Non-goals are load-bearing. Most spec disputes come from scope ambiguity. List what's explicitly out.
  • Task granularity. Each task in tasks.md should be completable in under ~2 hours and verifiable with a single command or check. If a task is "implement X feature," it's too big.
  • Link existing specs. Grep openspec/specs/ for related capabilities and cross-link them in design.md.

Anti-patterns:

  • Writing spec.md as a design doc. Specs describe behavior; design describes implementation.
  • Listing requirements as bullet points without scenarios. Bullet points are wishes; scenarios are testable.
  • Proposing a design with no alternatives. If you can't think of one, you haven't thought about it enough.
  • Copying the feature request verbatim into the proposal. Your job is to structure it, not to echo it.

Format and location:

Place new change folders under openspec/changes/<yyyy-mm-dd>-<slug>/. Use the current date (check with date if uncertain).

When done, summarize what you wrote and list any open questions that blocked you. Do not start implementation — that's /spec:apply.

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. 4d ago First seen · 37 lines · 56 tokens per session scan A 5fd074556d09

Subscribe to this mod's changes

spec-writer is an agent published in the GitHub repository orlando-japan/claude-code-setting (2 stars, last pushed 3mo ago), licensed MIT. It adds 56 tokens to every session and 567 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.