starchild-ai-agent/@4885/opensea

What this skill is for

Use this skill when the user asks for anything OpenSea-specific: NFT collection research, account portfolio views, listing/offer discovery, market events, token intel, or OpenSea trading workflow support.

Primary docs source:

• https://docs.opensea.io
• https://docs.opensea.io/llms.txt (index of docs + endpoint map)

API key strategy

OpenSea supports instant free-tier key creation (documented under Create an instant API key).

Use this decision rule:

1. If read-only request is small/occasional: create/use instant key.
2. If user wants sustained/production usage or higher throughput: request user-provided key via secure env input.

Never ask user to paste keys in chat.

Request pattern (Python scripts)

For any script you write for this skill:

• Use from core.http_client import proxied_get, proxied_post
• Add usage tracking header: headers={"SC-CALLER-ID": "chat:<thread_id>"} (or job/preview typed prefix when relevant)
• Read key from env if available (e.g., OPENSEA_API_KEY) and pass as X-API-KEY
• If key missing and task needs authenticated throughput, use request_env_input

Endpoint families to prioritize

Use OpenSea V2 reference endpoints grouped by intent:

1. Discovery / analytics

• collections: stats, traits, top/trending, holders, floor history
• NFTs: metadata, owners, analytics, by collection/account/contract
• events: global/by collection/by NFT/by account
• search: cross-entity search

2. Marketplace intelligence

• listings: best listing by NFT/collection, all listings by collection
• offers: best offer by NFT, collection offers, trait offers
• orders: get order, cancel order

3. Execution planning (no fabricated confirmations)

• listing/offer fulfillment data endpoints
• sweep and cross-chain fulfillment endpoints
• swaps: quote + execute transaction payloads
• transaction receipt endpoint for verification polling

Safety and verification

• Never claim a buy/list/swap/fulfillment is complete without receipt/status confirmation.
• Treat OpenSea as quote/action builder unless on-chain submission + receipt confirms execution.
• If endpoint output omits a hash or status, explicitly say it was not returned.

Common workflow templates

A) Collection research request

1. Resolve slug/collection.
2. Pull stats + floor history + traits.
3. Pull recent events.
4. Summarize liquidity, turnover, and near-term risk (washy volume, thin listings, trait concentration).

B) “Should I sweep this collection?”

1. Fetch best listings and recent sales events.
2. Compare spread vs recent clears.
3. Evaluate depth (how many items near floor).
4. If user wants action: generate sweep transaction plan and clearly label it as pending execution until receipt confirmed.

C) “Buy/list this NFT”

1. Pull best listing/offer + order details.
2. Generate fulfillment/listing actions.
3. Have wallet sign/broadcast via wallet tooling.
4. Poll transaction receipt endpoint and report verified state.

Known doc anchors

• API overview: /reference/api-overview
• API key flow: /reference/api-keys
• LLM/agent discovery: /reference/llms-agent-discovery
• SDKs: /reference/opensea-js, /reference/opensea-cli, /reference/stream-js

If endpoint behavior conflicts with assumptions, trust live docs and response payloads over memory.

Script usage

Script-mode skill — import from skills/opensea/exports.py.

python3 - <<'EOF'
import sys, json
sys.path.insert(0, "/data/workspace/skills/opensea")
from exports import os_collection_stats, os_search

print(json.dumps(os_collection_stats("doodles-official"), indent=2))
print(json.dumps(os_search("doodles", limit=3), indent=2))
EOF

Exports available

• API key: os_create_instant_api_key
• Collections: os_collections, os_collection, os_collection_stats, os_collection_holders, os_collection_floor_prices, os_collections_trending, os_collections_top
• Search: os_search
• NFTs: os_nft, os_nfts_by_contract
• Events: os_events, os_events_by_collection, os_events_by_nft
• Listings/Offers: os_listings_best, os_listings_all, os_offers_all, os_offers_for_nft, os_best_offer_for_nft
• Swap/verification: os_swap_quote, os_swap_execute, os_transaction_receipt (swap_quote required by OpenSea API; include at least one of transaction_identifiers / relay_request_id / request_id)

Ready scripts

• skills/opensea/scripts/smoke_test.py
• skills/opensea/scripts/get_collection_report.py
• skills/opensea/scripts/get_nft_snapshot.py
• skills/opensea/scripts/get_market_surface.py
• skills/opensea/scripts/poll_tx_receipt.py (expects swap_quote JSON inline or via @file.json)

End-to-end execution notes

1. os_swap_execute returns unsigned transaction payloads (to, data, value) and quote metadata; it does not broadcast.
2. Broadcast each transaction with wallet tooling (wallet_transfer) using chain-appropriate chain ID.
3. wallet_transfer may return user_operation_hash without a normal tx hash. You can use that value as transaction_identifiers[].transaction_hash for OpenSea receipt polling.
4. For receipt verification call os_transaction_receipt with:
   • a swap_quote object containing from_assets (and optionally to_assets), and
   • at least one identifier: transaction_identifiers (preferred), relay_request_id, or request_id.
5. request_id must be a valid UUID when supplied.
6. If swap_quote/identifier pairing is inconsistent, the API may return validation errors like provider-mapping failures.
7. Verified working pattern (Arbitrum USDC->WETH test):
   • swap_provider from quote: ZERO_EX
   • pass two transaction_identifiers with chain arbitrum, provider ZERO_EX, hashes from wallet user_operation_hash
   • receipt returns status: PENDING with populated asset_receipts/total_spent while settlement completes.

Output style

Return:

• user’s requested answer first,
• key metrics in compact bullets or table,
• explicit confidence/limits when data is partial,
• next actionable step (e.g. monitor, place order, wait for receipt).

No hype language, no fabricated execution results.