clarify

clarify is a skill for Claude Code, Codex from handarbeit/fabrik. It costs 124 tokens per session (3,303 once invoked), scanned A, original, Apache-2.0.

A specification-review tool that finds unclear or missing decisions in an existing feature description and asks targeted questions to resolve them.

In plain words
What is it for?
Use it to tighten a feature specification by asking up to five business-focused questions and writing each answer into the document.
Why use it?
It catches ambiguity before development begins, when answering a question is cheaper than rebuilding a feature later.

Skill for Claude CodeCodex

Part of the entwurf plugin — 6 skills shipped together

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 skills/handarbeit/fabrik/clarify
Any agent
npx skills add handarbeit/fabrik --skill clarify
Clone the repo
git clone --depth 1 https://github.com/handarbeit/fabrik

Made for: Claude Code, Codex.

Or install entwurf, the plugin that ships this one along with the rest of its 6 skills.

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 clarify

README.md
[![agentmods](https://agentmods.dev/badge/skills/handarbeit/fabrik/clarify.svg)](https://agentmods.dev/skills/handarbeit/fabrik/clarify)
Your own site
<a href="https://agentmods.dev/skills/handarbeit/fabrik/clarify"><img src="https://agentmods.dev/badge/skills/handarbeit/fabrik/clarify.svg" alt="Measured on agentmods" height="20"></a>
Per session 124 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,303 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 5.1 $0.00124 $0.03303
Opus 5 $0.00062 $0.01651
Sonnet 5 $0.00025 $0.00661
Haiku 4.5 $0.00012 $0.00330

Measured 5d ago against content hash f38733d8bd8f, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

clarify 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 5d 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.

plugin/entwurf/skills/clarify/SKILL.md · 156 lines

How it starts

The opening of the file, as written. The whole thing — 156 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Clarify a Specification

Find where a specification is ambiguous or silent, ask the smallest number of questions that resolve the highest-impact gaps, and write every answer back into the spec so the document — not the chat transcript — carries the decision.

This runs after a spec exists and before anyone builds against it. Ambiguity caught here costs one question; ambiguity caught during the build costs a rebuild.

The person answering is a product manager, business owner, or domain expert — expert in the business, not in software construction. Every question must be answerable from what they know about their own operation. A question they'd need an engineer to answer is a question you shouldn't be asking.

Never invent a specific number, threshold, name, or percentage they haven't given you. Turning a vague adjective into a number is the single most valuable thing this skill does — but the number has to come from them. Proposing "under two seconds" to someone with no target in mind produces agreement, not information, and a figure you made up is then indistinguishable in the spec from one the business actually committed to. Ask what they'd expect, or what they measure today; if they genuinely don't know, record the shape and leave the figure as [NEEDS BASELINE].

[NEEDS BASELINE] is a sanctioned marker meaning "not currently measured", written by this skill and by write-spec. It is the legacy spelling of [NEEDS MEASUREMENT], defined canonically in ../../CONVENTIONS.md; treat the two as identical and never rewrite an existing one. Resolve it only with a figure the person supplies. Never fill one in yourself, and never treat one as a lingering vague placeholder to be cleaned up.

Find the spec

Locate the specification in this order:

  1. If the person attached a spec to the conversation, work from that and hand back an updated copy when you're done. Everything below about reading and writing files applies to that working copy instead. Check this first — it costs nothing and saves hunting for a file that was handed to you.
  2. Otherwise establish which folder holds the spec: use the connected folder if there's exactly one, and ask which if several are connected. If no folder is connected, don't go looking in a temporary or scratch directory — ask them to attach the spec to the conversation instead, and tell them how to connect a folder so it's found automatically next time.
  3. Settle the project root: if a .specify/ folder exists at that folder or anywhere above it, the folder containing it is the root; otherwise the connected folder is. Everything below is relative to that root, and feature_directory is read relative to it — someone may have connected a sub-folder of a larger project whose .specify/ sits higher up.
  4. .specify/feature.json at the root — read feature_directory, and use <feature_directory>/spec.md. If the file can't be read, or the folder it names no longer exists, ignore it and fall through to the next step. Don't report that as an error — someone renaming a folder in Finder is a perfectly ordinary thing to do, and they have no idea a file was tracking it.
  5. If that's missing or stale, list the specs/NNN-* folders under the root and use the highest-numbered one, confirming with the person which spec they mean if there's any doubt. Once resolved, quietly update .specify/feature.json's feature_directory to point at the folder you settled on, preserving any other keys — leaving a stale pointer in place means every later run repeats this search, and it silently breaks write-spec's ability to revise a spec rather than starting a new one.
  6. If no spec file exists at all, don't create one — say plainly that there's no spec here yet and offer to write one with them first.

Read the full file on GitHub · 156 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. 5d ago First seen · 156 lines · 124 tokens per session scan A f38733d8bd8f

Subscribe to this mod's changes

clarify is a skill published in the GitHub repository handarbeit/fabrik (22 stars, last pushed 2d ago), licensed Apache-2.0. It adds 124 tokens to every session and 3,303 once invoked, about $0.0006 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 2026-08-30.