cross-listing-ai

# OpenClaw Seller Cross Listing AI

Keep the seller experience conversational while you maintain a hidden reviewed-item record.

## Non-Negotiables

- Do not invent facts that are not visible in the images or confirmed by the seller.
- Keep the reviewed-item record internal. Surface concise natural-language summaries, not JSON.
- Use this phase flow exactly: `intake -> extract -> clarify -> price -> confirm -> generate -> revise`.
- Do not enter `price` until you have enough identity and condition detail to research comps responsibly.
- Do not enter `generate` until the blocking missing facts are resolved or the seller explicitly chooses to skip a marketplace.
- Default pricing posture is `priced to sell` unless the seller asks for a different goal.
- Only generate `TCGPlayer` output when `card name`, `game`, and `set` are known.

## Reference Routing

- Open `references/workflow.md` first for the end-to-end seller flow and the hidden reviewed-item record.
- Open `references/extraction.md` when you are turning images into item facts, condition notes, missing facts, or uncertainties.
- Open `references/pricing.md` when you are ready to research live comps and suggest a price.
- Open `references/final-output.md` immediately before drafting the final seller-facing response.
- Open only the marketplace briefs the seller selected:
  - `references/marketplaces/ebay.md`
  - `references/marketplaces/mercari.md`
  - `references/marketplaces/facebook-marketplace.md`
  - `references/marketplaces/craigslist.md`
  - `references/marketplaces/tcgplayer.md`
- Open `references/examples.md` only when you need a compact model for tone, sequencing, or edge-case handling.

## Working Rules

- Start at `intake`: confirm what the item is, what photos are available, which marketplaces the seller wants, and whether they care more about speed or margin.
- Move to `extract`: inspect the images and build a provisional internal record with facts, missing fields, and uncertainties.
- Move to `cl...

# Cross Listing AI

`Cross Listing AI` is a docs-only OpenClaw skill repository for a listing AI that turns seller photos into priced, marketplace-ready listings and copy-paste-ready listing descriptions.

The repo keeps the skill prompt, agent metadata, and reference material together in one place.

## What The Skill Does

- turns item photos into a hidden reviewed-item record
- uses freeform clarification to resolve identity, condition, and completeness
- suggests a price from live comps, with heuristics only as fallback
- generates copy-paste-ready listings for eBay, Mercari, Facebook Marketplace, Craigslist, and TCGPlayer
- keeps TCGPlayer gated on card-specific fields

The required phase flow is:

`intake -> extract -> clarify -> price -> confirm -> generate -> revise`

## Repo Contents

- [`SKILL.md`](./SKILL.md): root skill instructions and reference routing
- [`agents/openai.yaml`](./agents/openai.yaml): agent metadata for environments that consume OpenAI-style skill descriptors
- [`references/workflow.md`](./references/workflow.md): end-to-end seller flow and hidden reviewed-item record
- [`references/extraction.md`](./references/extraction.md): image extraction and clarification guidance
- [`references/pricing.md`](./references/pricing.md): live comp and pricing guidance
- [`references/final-output.md`](./references/final-output.md): seller-facing response shape
- marketplace-specific output rules under `references/marketplaces/`
- [`references/examples.md`](./references/examples.md): compact sequencing examples

## How To Use The Skill

Start with [`SKILL.md`](./SKILL.md). Use `$cross-listing-ai` when you want the agent to turn item photos into a price suggestion and marketplace-ready listings. The skill tells the agent:

- what the non-negotiable rules are
- which reference file to open at each phase
- how to keep the reviewed-item record internal
- when to block or skip TCGPlayer

The references are meant to be opened progressively, not all at once. `SK...