메뉴
Automated daily swing trading strategy — scans pre-market movers, picks momentum stocks, buys at open, manages exits at afternoon check with stop-loss and profit targets. Designed for small cash accounts ($500) using Robinhood Agentic MCP.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | ada-trader |
| description | Automated daily swing trading strategy — scans pre-market movers, picks momentum stocks, buys at open, manages exits at afternoon check with stop-loss and profit targets. Designed for small cash accounts ($500) using Robinhood Agentic MCP. |
| metadata | {"author":"github.com/LeoSaucedo","version":"1.0.0","openclaw":{"emoji":"📈","requires":{"bins":["node","pnpm","python3"],"skills":["robinhood-agentic"]}}} |
Daily swing trading with dynamic stock selection. Scans pre-market movers, ranks by momentum, buys at open, manages exits with stop-loss and profit targets.
The Robinhood Agentic skill must be installed and authenticated:
cd robinhood-agentic && pnpm install && node rh-client.mjs auth
Two crons required, both pointing at this skill:
Morning Routine (M-F, 9:30 AM ET):
schedule: "30 9 * * 1-5" (America/New_York)
payload: "Read the ada-trader skill and execute the Morning Routine. Post to your trading channel."
delivery: announce to your trading channel
timeout: 600s, thinking: medium
Afternoon Routine (M-F, 3:45 PM ET):
schedule: "45 15 * * 1-5" (America/New_York)
payload: "Read the ada-trader skill and execute the Afternoon Routine. Post to your trading channel."
delivery: announce to your trading channel
timeout: 600s, thinking: minimal
Adjust max positions in the Strategy Parameters table. Trade size is automatically set to 10% of total account value — no manual editing needed.
| Parameter | Value |
|---|---|
| Chunk size | 10% of account value (auto-calculated) |
| Max positions | 5 (or until settled cash exhausted) |
| Stop-loss | -5% from entry |
| Profit target | +10% from entry |
| Daily sell rule | Close the oldest open position every afternoon |
Pick fresh tickers each morning — no static watchlist. Use web_search for:
Compile 10-15 tickers with an absolute move of at least ±0.5% in pre-market. Prefer stocks with clear catalysts (earnings, analyst upgrades, sector momentum, news). Flat market = fewer picks.
Search for pre-market movers and compile 10-15 tickers.
echo "TICKER1,TICKER2,..." | python3 ~/.openclaw/workspace/skills/ada-trader/scripts/scanner.py
Pick the top-ranked symbol. The scanner calls the Robinhood MCP for live quotes and ranks by momentum (% change from prior close).
Before announcing a pick, check if the top-ranked symbol (or any candidate) was recently sold. Run:
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call get_equity_orders "{\"account_number\":\"<ACCT>\",\"side\":\"sell\",\"state\":\"filled\",\"created_at_gte\":\"$(date -u -d '24 hours ago' +%Y-%m-%dT%H:%M:%SZ)\"}"
Extract symbols from filled sell orders in the last 24 hours. Skip any ticker that was sold within the last 24 hours — move to the next-ranked candidate instead. Post to channel: "Skipping {TICKER} — sold within last 24h." If all candidates are blacklisted, skip trading for the day and post why.
Post to channel: symbol, current price, % change, catalyst, target +10%, stop -5%.
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call get_portfolio '{"account_number":"<ACCT>"}'
Extract total_value. Calculate chunk = total_value × 0.10 (rounded to 2 decimals). Verify settled cash ≥ chunk. Get account_number from get_accounts if needed. If not enough cash, post why and skip.
First, review the order to surface pre-trade warnings:
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call review_equity_order '{"account_number":"<ACCT>","symbol":"<PICK>","side":"buy","type":"market","dollar_amount":"<CHUNK>","time_in_force":"gfd","market_hours":"regular_hours"}'
Post any alerts from the review output to the channel alongside the estimated cost and quote data.
⚠️ place_equity_order executes immediately — there is no Robinhood app confirmation step. The review step is the only preview; once placed, the order is sent directly.
Then place the order:
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call place_equity_order '{"account_number":"<ACCT>","symbol":"<PICK>","side":"buy","type":"market","dollar_amount":"<CHUNK>","time_in_force":"gfd","market_hours":"regular_hours"}'
Replace <CHUNK> with the calculated 10% value (e.g. "50.00" for a $500 account).
Advanced users may skip the review step if they understand the risks — but still get chat confirmation before placing.
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call get_equity_positions '{"account_number":"<ACCT>"}'
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call get_equity_quotes '{"symbols":["TICKER1","TICKER2"]}'
Calculate P&L: (current_price - avg_cost) / avg_cost × 100.
First, review the sell order to surface any pre-trade warnings:
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call review_equity_order '{"account_number":"<ACCT>","symbol":"<TICKER>","side":"sell","type":"market","quantity":"<ALL SHARES FROM POSITIONS>","time_in_force":"gfd","market_hours":"regular_hours"}'
Post any alerts alongside the P&L and current quote.
⚠️ place_equity_order executes immediately — there is no Robinhood app confirmation step. The review step is the only preview; once placed, the order is sent directly.
Then place:
node ~/.openclaw/workspace/skills/robinhood-agentic/rh-client.mjs call place_equity_order '{"account_number":"<ACCT>","symbol":"<TICKER>","side":"sell","type":"market","quantity":"<ALL SHARES FROM POSITIONS>","time_in_force":"gfd","market_hours":"regular_hours"}'
Use quantity (number of shares), NOT dollar_amount, for sell orders.
Advanced users may skip the review step if they understand the risks — but still get chat confirmation before placing.
Post to channel: what sold, P&L per position, remaining positions.
Connect to Robinhood Agentic Trading via MCP — view portfolio, analyze positions, place trades, and execute automated strategies through Robinhood's official MCP server.
Interact with SoundCloud API for searching tracks, managing playlists, user operations, and audio discovery. Use when the user asks to search for music on SoundCloud, create/edit playlists, get track information, find user profiles, manage favorites, or work with SoundCloud audio content. Requires SOUNDCLOUD_CLIENT_ID and SOUNDCLOUD_CLIENT_SECRET for all operations; SOUNDCLOUD_USER_TOKEN for write operations (playlists, likes, follows).
Gmail inbox triage with self-learning SQLite pattern database. Multi-account support with whitelist/blacklist, keyword-domain-essence scoring, and daily feedback sweep learning. Use when the user asks to check email, triage inbox, or during automated email triage cron runs.
Graph-based agent memory for tracing relationships between facts. Use when memory_search returns isolated facts and you need to see how they connect — causal chains, hierarchies, related events. LTM-only, stored as JSON graph with typed nodes and weighted edges.
Query WaniKani Japanese learning data — profile, kanji/vocab lookup, review forecast, SRS progress, and study statistics. Requires a WaniKani API token.
Check DeepSeek API balance and usage. Use when the user asks about - "Check my DeepSeek balance", "How much API credit do I have left?", "What's my DeepSeek usage?", "DeepSeek account status", "API token balance", or any query related to DeepSeek platform credits, usage, or account information.