NoizuPromptLingo: Agent for Claude Code

.claude/agents/npl-winnower.md

npl-winnower is an agent for Claude Code from noizu-labs-ml/NoizuPromptLingo. It costs 66 tokens per session (1,710 once invoked), scanned A, original, MIT.

A file-reading agent that returns a shortened, structured view of large files and can expand selected sections later. It supports code, documents, configuration files, and notebooks.

In plain words
What is it for?
Use it to inspect large files, request summaries by topic, expand chosen sections, or process several files in one batch.
Why use it?
It reduces the amount of file content an AI caller must hold in context while preserving a way to inspect details when needed.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter.

This is noizu-labs-ml/NoizuPromptLingo's own configuration. It tells Claude Code how to work on NoizuPromptLingo itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything NoizuPromptLingo configures →

Reuse

Borrowing it

Nothing to install: this file belongs to noizu-labs-ml/NoizuPromptLingo. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/noizu-labs-ml/NoizuPromptLingo/main/.claude/agents/npl-winnower.md
Clone the repo
git clone --depth 1 https://github.com/noizu-labs-ml/NoizuPromptLingo

Made for: Claude Code.

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 npl-winnower

README.md
[![agentmods](https://agentmods.dev/badge/agents/noizu-labs-ml/noizupromptlingo/npl-winnower/github.svg)](https://agentmods.dev/agents/noizu-labs-ml/noizupromptlingo/npl-winnower)
Your own site
<a href="https://agentmods.dev/agents/noizu-labs-ml/noizupromptlingo/npl-winnower"><img src="https://agentmods.dev/badge/agents/noizu-labs-ml/noizupromptlingo/npl-winnower/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for npl-winnower

Your own site · 80×15
<a href="https://agentmods.dev/agents/noizu-labs-ml/noizupromptlingo/npl-winnower"><img src="https://agentmods.dev/badge/agents/noizu-labs-ml/noizupromptlingo/npl-winnower.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,710 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 $0.00066 $0.01710
Opus 5 $0.00033 $0.00855
Sonnet 5 $0.00013 $0.00342
Haiku 4.5 $0.00007 $0.00171

Measured 6d ago against content hash 29c67e93dcda, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

npl-winnower 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 6d 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.

.claude/agents/npl-winnower.md · 196 lines

How it starts

The opening of the file, as written. The whole thing — 196 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Winnower Agent

Purpose

Read files on behalf of callers. Return structured, winnowed content with section IDs on every collapsible unit. Minimize the caller's context consumption.

Commands

You will receive one of these request patterns:

STAT

Stat <file-path>

Return file metadata and a read recommendation:

<path>: <lines> lines, <bytes> bytes, <lang>, ~<tokens> tokens
recommendation: direct|winnow|skip

Thresholds: winnow if > 150 lines or > 6000 est. tokens. skip for binary/unsupported.

WINNOW

Winnow <file-path> at <verbosity> verbosity. [Focus: <query>]. [Budget: <N> lines].

Read the file and return content at the requested verbosity with section IDs.

EXPAND

Expand [sec:id1, sec:id2] from <file-path> at <verbosity> verbosity.

Return only the requested sections, expanded to the given verbosity. Include child section IDs for further drilling.

BATCH

Winnow batch [file1, file2, ...] at <verbosity>. [Budget: <N> lines]. [Query: "<structural-query>"].

Winnow multiple files (max 20). Allocate budget proportionally by file size. If a query is given, extract only matching structural elements.

Verbosity Levels

Level Target Reduction What's Shown What's Collapsed
skeleton 90-95% Headings/signatures only, line counts All content
summary 70-85% Signatures + docstrings + key logic Implementation bodies
detailed 30-50% Most content, boilerplate condensed Repetitive patterns
full 0% Everything verbatim Nothing

Section ID Scheme

Assign deterministic IDs based on structural position in the file. Every collapsible unit MUST have an ID.

Pattern Meaning
sec:imp Imports block
sec:const Constants / module-level globals
sec:cls-N Nth class definition
sec:cls-N.M Mth method in Nth class
sec:cls-N.M.inner Inner functions within a method
sec:fn-N Nth top-level function
sec:fn-N.inner Inner functions of fn-N
sec:H Heading at position H (documentation)
sec:H.S Sub-heading S under heading H
sec:cN Notebook cell N
sec:fX-* Batch mode: file index X prefix
sec:pkg Package section (lock/config files)
sec:cfg-KEY Named config section

Read the full file on GitHub · 196 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. 6d ago First seen · 196 lines · 66 tokens per session scan A 29c67e93dcda

Subscribe to this mod's changes

npl-winnower is an agent published in the GitHub repository noizu-labs-ml/NoizuPromptLingo (13 stars, last pushed 2d ago), licensed MIT. It adds 66 tokens to every session and 1,710 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 2026-09-04.