xianyu-product-ops

xianyu-product-ops

This is a composite operations skill for Xianyu; the core object is the product, not the content. It combines xianyu-explore、xianyu-publish and xianyu-interact to complete the operations loop.

Applicable scenarios

• The user requests analysis of whether a certain Xianyu category has demand.
• The user requests publishing a product to test traffic after referencing competitors.
• The user requests optimizing the title, description, price, and image selling points.
• The user requests reviewing views, wants, consultations, or transaction leads.
• The user requests assistance in replying to buyer inquiries.

Common commands

python scripts/cli.py analyze-competitors --keyword "PPT代做" --limit 10
python scripts/cli.py search-items --keyword "关键词" --limit 20
python scripts/cli.py get-item-detail --item-id ITEM_ID
python scripts/cli.py fill-publish --description-file D:/abs/desc.txt --price 29.9 --images D:/abs/cover.png
python scripts/cli.py click-publish
python scripts/cli.py get-item-stats --item-id ITEM_ID
python scripts/cli.py list-my-items
python scripts/cli.py list-messages

Competitor analysis process

1. Clarify keywords; prioritize covering category words, scenario words, and transaction words at the same time.
2. Run analyze-competitors or search-items。
3. Select products with high want counts, clear titles, and explicit prices.
4. Run get-item-detail。
5. for key products. Output price bands, high-frequency title words, service scenarios, commitment methods, image directions, and potential risks.

During analysis, prioritize extracting:

• Price range and median.
• Title structure of high-want products.
• Service boundaries, delivery cycle, and number of revisions in the product description.
• Whether the images highlight the scenario, result, price, or service process.
• Platform risk warnings, such as academic ghostwriting, external traffic diversion, and exaggerated promises.

Publishing test process

1. Do competitor search first; do not create titles and prices behind closed doors.
2. Use the official image materials provided by the user; if the user provides image URLs, download them as local files first.
3. Generate or organize the description file; avoid contact information, QR codes, and off-platform traffic diversion.
4. Use fill-publish to fill in the form step by step.
5. Check category_resolution、unsupported_category_marker and selected。
6. After user confirmation, execute click-publish, or execute publish。
7. when the user explicitly authorizes it. Record the product ID, URL, price, final category, and publishing time.

Traffic review process

1. Use get-item-stats --item-id ITEM_ID to read views and wants.
2. Use list-messages to check whether consultations have appeared.
3. Compare publishing time, price, title, images, and competitor data.
4. Give next-step adjustment suggestions: title, first image, price, description, category, or delist and republish.

During review, do not look only at view count. What is more valuable is:

• Views -> Wants
• Wants -> Consultations
• Consultations -> Clear demand
• Clear demand -> Transaction possibility

Consultation assistance process

1. Use list-messages to view the consultation summary.
2. Summarize buyer questions: price, delivery time, number of revisions, material requirements, and whether it is urgent.
3. Generate a reply draft.
4. After user confirmation, use send-message to send.

Reply suggestions:

• Confirm the requirements first; do not directly promise guaranteed passing or guaranteed results.
• Ask about purpose, page count, deadline, existing materials, and style preferences.
• For service-type products, prioritize phrasing the quote as "confirmed based on page count and complexity."
• Avoid off-platform traffic diversion and sensitive commitments.

Strategy Principles

• Expose full functionality; the documentation layer indicates confirmation and frequency, while the CLI layer does not impose hard restrictions.
• High-impact actions such as publishing, changing prices, direct messaging, deleting, and delisting should be confirmed first.
• Operations analysis should be based on real searches and product details first, not just experience-based guesses.
• For official publishing, prioritize using the user's real materials, and do not treat ImageGen as a required capability across Agents.

Failure Handling

• Few search results: switch keywords, use broader category terms or scenario terms.
• Abnormal competitor data: search again, and avoid treating ads, irrelevant products, or recommendation section data as samples.
• Product delisted after a publishing test: check message notifications and product details, and adjust the description, images, or category.
• Insufficient message parameters: first obtain them through product details or the page link itemId / peerUserId。