# AI Writing Audit

- kind: Eval
- sku: sku_ai_writing_audit
- version: 1.0.0
- price: $15 USD
- availability: in_stock
- runtime: Claude Code, Codex, Cowork, Grok, OpenClaw
- last_tested: 2026-08-22
- page: https://kitsforagents.com/kits/evals/ai-writing-audit/
- json: https://kitsforagents.com/kits/evals/ai-writing-audit.json

## Job

Gate text a human will put their name on: tag LLM-style tells, rewrite, and leave a changelog.

## Who it is for

A content agent that publishes under a human name.

## Who should skip it

You only need the free public checklist on GitHub.

## What you get

Installable eval skill that tags common LLM-style tells, rewrites the passage, and writes a changelog. Includes fixtures that must fail on em dashes, AI lexicon, and rule-of-three stacks. Public repo remains free; this is the gated folder with evals.

## Inputs

- draft (text, required): The draft to audit

## Outputs

- audit (markdown): Tagged audit
- revised (markdown): Corrected text
- changelog (markdown): What changed and why

## Eval

100/100 · 2/2 · 2026-08-22 · Public a-makelky/ai-writing-audit fixtures

Fail fixture tags; clean fixture passes.

## Success criteria

- Fixture with em-dash, AI-lex, and rule-of-3 emits those tags
- Corrected text no longer contains the tagged tells
- Changelog lists each change

## License and refund

Personal-use, one seat, no redistribution. 14-day refund if the advertised eval fails on the advertised fixture, or the checksum mismatches.

## Checkout

Human: POST https://kitsforagents.com/api/v1/checkout with `{ "sku": "sku_ai_writing_audit" }`
Agent: POST https://kitsforagents.com/api/acp/checkout_sessions with ACP 2026-04-17 checkout session body.

```yaml
schema: "makelky.agent-packet/2026-08"
sku: "sku_ai_writing_audit"
name: "AI Writing Audit"
kind: "evals"
version: "1.0.0"
url: "https://kitsforagents.com/kits/evals/ai-writing-audit/"
md_url: "https://kitsforagents.com/kits/evals/ai-writing-audit.md"
json_url: "https://kitsforagents.com/kits/evals/ai-writing-audit.json"
job: >
    Gate text a human will put their name on: tag LLM-style tells, rewrite, and leave a changelog.
summary: >
    A style audit, not authorship detection. The public GitHub skill stays the lead magnet. This pack adds eval fixtures and a version pin.
inputs:
  - name: "draft"
    type: "text"
    required: true
    description: "The draft to audit"
outputs:
  - name: "audit"
    type: "markdown"
    description: "Tagged audit"
  - name: "revised"
    type: "markdown"
    description: "Corrected text"
  - name: "changelog"
    type: "markdown"
    description: "What changed and why"
runtime:
  primary: "claude-code"
  also:
    - "codex"
    - "cowork"
    - "grok"
    - "openclaw"
permissions: "read the draft; write audit files"
secrets_required: []
eval:
  score: 100
  casesPassed: 2
  casesTotal: 2
  dateTested: "2026-08-22"
  environment: "Public a-makelky/ai-writing-audit fixtures"
  summary: "Fail fixture tags; clean fixture passes."
price:
  amount: 15
  amount_minor: 1500
  currency: "USD"
  display: "$15"
  is_digital: true
  availability: "in_stock"
license:
  id: "personal-use"
  url: "https://kitsforagents.com/legal/license/"
  redistribution: "forbidden"
  seats: 1
  commercial: false
install:
  after_purchase: true
  preview_url: "https://kitsforagents.com/install/ai-writing-audit/"
success_criteria:
  - "Fixture with em-dash, AI-lex, and rule-of-3 emits those tags"
  - "Corrected text no longer contains the tagged tells"
  - "Changelog lists each change"
refund:
  window_days: 14
  policy_url: "https://kitsforagents.com/refunds/"
  agent_endpoint: "https://kitsforagents.com/api/v1/refunds"
  trigger: "success_criteria_unmet_or_install_checksum_mismatch"
budget:
  class: "low"
risk:
  level: "low"
author:
  name: "Aaron Makelky"
  url: "https://aaronmakelky.com/about"
  tested_on_own_work: true
  brand_deals: false
checkout:
  human: "https://kitsforagents.com/api/v1/checkout"
  acp: "https://kitsforagents.com/api/acp/checkout_sessions"
last_tested: "2026-08-22"
is_eligible_checkout: true
```