# Red String

> Plan a whodunit before you draft it. Pin the suspects and the clues of a mystery outline to an evidence corkboard, chapter by chapter, and see which clues the reader actually gets before the detective names the culprit. A metered fair-play reviewer then calls the book fair, thin or cheats, names the suspect a careful reader would guess, rates every clue and proposes plants by chapter.

URL: https://red-string.skillsafe.ai/

## What is free (runs in the browser)

- The case: chapters (3-80), the reveal chapter, a working title and the premise with the solution.
- Suspects (up to 8): name, role, motive, the chapter where each first appears, and which one is the culprit.
- Clues (up to 30): what the reader is shown, the chapter, the suspect it points at, and on or off the page.
- The board: an SVG corkboard with a string from every clue to its suspect, tally marks of on-page clues on every suspect card, the reveal line and the first third of the book shaded. A spoiler-safe mode for sharing with a beta reader draws every string in one colour, hides the culprit stamp, the counts and the checks, and makes the .svg, .md and .csv exports leave out the culprit, the premise and the review.
- Fair-play checks: every clue against the culprit sorted into earnable, off the page or at/after the reveal; red herrings; a late culprit (after the first third); every real clue crammed into the last quarter; suspects with no on-page clue.
- Text import (one line per suspect or clue), .txt/.json file drop, exports: Markdown case file, clues CSV, case JSON, board SVG, a printable board.

The thresholds are genre rules of thumb, not laws.

## What is metered

One task, `audit`, billed in SkillSafe credits to a signed-in account. Input: `task`, `title`, `premise`, `chapters`, `reveal`, `culprit`, `suspects`, `clues`, `facts` (JSON-encoded). Reply: one JSON object with `verdict` (`fair` | `thin` | `cheats`), `verdict_why`, `title`, `reader_guess` {id, why, cite}, `suspects` [{id, suspicion 0-5, note, cite}], `clues` [{id, strength: strong|weak|none, note, cite}], `plants` [{chapter, points, text, why, cite}], `tips`, `caveats`. The page checks every reply against the board before its plants can be pinned.

API docs: https://red-string.skillsafe.ai/api.html

## Source

The reviewer is built on the SkillSafe skill @wordflowlab/mystery-novel-conventions (https://skillsafe.ai/skill/@wordflowlab/mystery-novel-conventions): mystery, detective and crime genre conventions, clue placement and fair-play principles.