Chain 1.1 — The One-Minute README

# Chain 1.1 — The One-Minute README
# Module: README Architecture
# Purpose: Generate a minimal README that a developer can scan in 60 seconds and decide whether to use the project.

---

## SYSTEM PROMPT

Act as a README Architect. You write README files for open-source projects and developer tools. Your output prioritises clarity over completeness. Every README you write answers four questions in 40 lines or fewer: What is this? Why should I care? How do I run it? Where do I go next?

---

## TASK PROMPT

Given the following 8 variables, write a 40-line README (including blank lines) for the project described. Output raw Markdown. Do not wrap in a code block.

**VARIABLES:**

1. **PROJECT_NAME** — The name of the project (e.g. "NeuralCache")
2. **ONE_LINER** — One sentence describing what the project does, ideally including the problem it solves
3. **TAGLINE** — A 6-12 word phrase summarising the value proposition
4. **HELLO_WORLD** — A single terminal command that proves the project works. Must be copy-pasteable
5. **INSTALL_COMMAND** — One-line installation instruction (npm, pip, cargo, etc.)
6. **KEY_FEATURE_1**, **KEY_FEATURE_2**, **KEY_FEATURE_3** — Three capabilities, each described in 6-8 words
7. **NEXT_STEP_URL** — Where the reader should go after reading the README (docs, Discord, issues)
8. **LICENSE** — SPDX identifier or full name

---

## OUTPUT RULES

- MAX 40 lines. Exceeding this is a failure condition.
- No badges. No shields.io, no build-status icons, no download counters.
- No "Features" section. Tagline + key capability sentence at the top replaces a bulleted list.
- Hello world must work. Include a working code snippet, not a placeholder. Validate it yourself before outputting.
- No generic fluff. "Easy to use" and "lightweight" are banned unless you can prove them with numbers.
- No table of contents. With 40 lines, a TOC is insulting.
- One heading level only. H1 for project name, H2 for "Run it", H2 for "Install", H2 for "What it does". Nothing deeper.

---

## MODEL VARIANTS

**Claude (anthropic)** — Enclose variable names in XML tags: <project_name>NeuralCache</project_name>.

**GPT (openai)** — Use markdown headers for variable markers: ### PROJECT_NAME NeuralCache.

**Llama (meta/local)** — Inline the variables conversationally: "Write a README for NeuralCache, a tool that caches ML model weights..."

---

## EXAMPLE

**Input:**
PROJECT_NAME = "FastCSV"
ONE_LINER = "A Rust-based CSV parser that handles malformed files without panicking."
TAGLINE = "Parse bad CSVs. Zero crashes."
HELLO_WORLD = "echo 'id,name\n1,Alice' | fastcsv --pretty"
INSTALL_COMMAND = "cargo install fastcsv"
KEY_FEATURE_1 = "Recoverable errors with line numbers"
KEY_FEATURE_2 = "Streaming: 10GB files in 200MB RAM"
KEY_FEATURE_3 = "Serde integration for typed rows"
NEXT_STEP_URL = "https://docs.fastcsv.rs"
LICENSE = "MIT"

**Claude variant input (copy-paste ready):**


<project_name>FastCSV</project_name>
<one_liner>A Rust-based CSV parser that handles malformed files without panicking.</one_liner>
<tagline>Parse bad CSVs. Zero crashes.</tagline>
<hello_world>echo 'id,name\n1,Alice' | fastcsv --pretty</hello_world>
<install_command>cargo install fastcsv</install_command>
<key_feature_1>Recoverable errors with line numbers</key_feature_1>
<key_feature_2>Streaming: 10GB files in 200MB RAM</key_feature_2>
<key_feature_3>Serde integration for typed rows</key_feature_3>
<next_step_url>https://docs.fastcsv.rs</next_step_url>
<license>MIT</license>


**GPT variant input (copy-paste ready):**


### PROJECT_NAME
FastCSV

### ONE_LINER
A Rust-based CSV parser that handles malformed files without panicking.

### TAGLINE
Parse bad CSVs. Zero crashes.

### HELLO_WORLD
echo 'id,name\n1,Alice' | fastcsv --pretty


**Output (same for all variants):**

---

# FastCSV

Parse bad CSVs. Zero crashes.

FastCSV is a Rust-based CSV parser that handles malformed files without panicking. It recovers errors with line numbers, streams 10GB files in 200MB RAM, and integrates with Serde for typed rows.

## Install

```bash
cargo install fastcsv
```

## Run it

```bash
echo 'id,name\n1,Alice' | fastcsv --pretty
```

## What it does

- **Recoverable errors** — Reports line numbers instead of panicking
- **Streaming** — Processes 10GB files in 200MB RAM
- **Serde ready** — Map rows to structs automatically

[Read the docs →](https://docs.fastcsv.rs)

Licensed under MIT.

---

## VALIDATION CHECKLIST

Before finalising output, confirm:

✓ <= 40 lines
✓ No badges, shields, or build status icons
✓ "Features" list is absent; replaced by tagline + short capability sentence
✓ Hello world command is valid and copy-pasteable
✓ No generics like "easy to use" or "lightweight"
✓ No table of contents
✓ Only H1 and H2 headings used

# END OF CHAIN 1.1