Guide
AI Prompt Cookbook for TradingLead (40+ Examples)
This cookbook lists more than 40 realistic requests you can give an AI assistant connected to TradingLead Desktop. Examples cover charts, drawings, watchlists, screening, alerts, Pine and Python indicators, backtests, walk-forward optimization, paper trading, bar replay, Historical Patterns, research pages and demo recording. Each row names the public skill and the main commands the assistant should use, and what you get back. Every example is a simulation or research task; none places a live order.
How to use this cookbook
Each table row is a request you can paste into GitHub Copilot CLI, Claude Code or another terminal agent connected to TradingLead Desktop. Replace the symbols, intervals and numbers with your own. The Skill column names the public skill that owns the task. Main commands are real names from the public command index, described on TradingLead commands for AI agents. You get back is what the assistant should report.
Start every session with one line:
Run
tradinglead agent-guide, list the open TradingLead tabs, target my active chart tab and read its capabilities before doing anything.
The assistant should then check the account, symbol, feed and interval (chart.getState, workspace.getState). Setup details are in control TradingLead with AI.
Charts and multi-timeframe analysis
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Open NSE:TCS-EQ on the daily chart with the default feed." | tl-charts | feeds.list, chart.open, chart.getState | The chart switches; the assistant confirms the symbol, feed, interval and that data is ready |
| "Show the last 20 daily candles of the current chart as a table." | tl-charts | chart.snapshot | Up to 25 closed candles with timestamps, with any gaps noted |
| "Switch to Heikin Ashi, log scale and a 6-month range." | tl-charts | chart.viewState, chart.setView | The new chart type, scale and visible range, read back from the chart |
| "Compare INFY against TCS as a percentage overlay." | tl-charts | chart.compare | A Compare overlay on the active chart |
| "Split into two panes: daily on the left, 15-minute on the right, same symbol." | tl-charts | layout.set, chart.open (per active pane), layout.getState | A two-pane layout for multi-timeframe analysis |
| "Save this layout as 'Swing MTF' and reload it later." | tl-charts | layout.saved | A saved chart layout you can load again |
| "Open the stocks from my last scan as Chart Grid cards on 1D." | tl-charts | grid.open, grid.configure | A cloud-backed Chart Grid of the selected tickers |
| "Add the built-in RSI and a 50-period EMA, then hide the EMA." | tl-charts | studies.list, studies.apply, studies.setVisibility | Native studies on the chart, with their IDs |
Drawings and templates
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Draw a trend line between the March low and the June low." | tl-charts | chart.snapshot or marketData.getBars, drawings.create | One trend line placed from real candle times and prices |
| "List my drawings on this chart and lock the horizontal lines." | tl-charts | drawings.list, drawings.update | Drawing IDs and their new lock state |
| "Save my current indicators as a template called 'Momentum'." | tl-charts | studyTemplates.manage | A saved native indicator template |
See drawing tools. Trendline alerts still have to be drawn by you.
Watchlists and market panels
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Read my 'Swing' watchlist sorted by % change and give me CSV text." | tl-markets | watchlists.select, watchlists.getRows (format csv) | Up to 100 rows per page as formula-safe CSV text; nothing written to a file |
| "Create a watchlist 'Breakouts Nov' with RELIANCE, TCS and INFY." | tl-markets | watchlists.create, watchlists.addSymbols | A new cloud watchlist (needs a real sign-in) |
| "Show peers for the current stock and refresh them." | tl-markets | workspace.setPanel, peers.get, peers.configure | The peers table from the bottom panel |
| "Get the latest NSE participant futures report for FII positions." | tl-markets | workspace.setPanel, futures.getReport | Published participant open-interest data, as shown in futures analysis |
Screening
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Scan NIFTY 50 stocks on 1D for close above SMA20 and RSI14 ≥ 60." | tl-screening | screening.catalog, screening.listUniverses, screening.configure, screening.run, screening.getResults | Matches with warnings and coverage counts, 10–25 rows per page |
| "From those matches, keep only the ones I selected in the grid and rescan them with RSI14 above 65." | tl-screening | screening.refine, screening.run | A follow-up scan of only the earlier matches |
| "Save this rule as 'RSI momentum' so I can rerun it." | tl-screening | screening.save, screening.listSaved | A saved screener |
| "Reopen yesterday's scan result without rerunning it." | tl-screening | screening.listHistory, screening.openHistory | The stored snapshot, with no new allowance used |
| "Add every match to my 'Watch next' list." | tl-screening + tl-markets | screening.getResults (all pages), watchlists.addSymbols | Symbols added to a TradingLead (not broker) watchlist |
| "Run my Pine indicator on 30 tickers using my connected broker's candles." | tl-screening | providerScreener.listCapabilities, providerScreener.run | A local broker scan of 1–50 tickers; results expire after 30 minutes |
Walkthrough: screen stocks with an AI assistant. Feature page: stock screener.
Alerts
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Create an alert when NSE:TCS-EQ crosses above 4,000, once." | tl-alerts | alerts.create, alerts.list | A saved price-alert rule |
| "Pause all my alerts on INFY." | tl-alerts | alerts.list, alerts.setEnabled | Each named rule paused, read back |
| "Show which alerts triggered this week." | tl-alerts | alerts.activity | Persisted trigger events, page by page |
Alerts are checked only while the Alerts panel is open in a signed-in tab. There is no push, SMS or email; see alerts.
Pine and Python indicators
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Write a Pine v5 indicator plotting a 20/50 EMA ribbon, validate it and add it to my chart." | tl-tlscript | indicators.listCapabilities, indicators.validate, indicators.apply | Diagnostics, then an applied study |
| "Save that indicator to my library and change the fast length to 13." | tl-tlscript | indicators.save, indicators.update | A saved revision and an updated chart instance |
| "Port this Pine script and tell me what isn't supported." | tl-tlscript | indicators.validate | Explicit diagnostics for unsupported constructs |
| "Write a Python indicator that marks gap-ups over 2% and apply it." | tl-python | pythonIndicators.capabilities, pythonIndicators.validate, pythonIndicators.apply | A syntax check, then plots on the chart |
| "Restore the previous version of my Python indicator." | tl-python | pythonIndicators.history, pythonIndicators.restore | The earlier source revision |
Walkthrough: build an indicator with an AI assistant. See Pine Script and Python indicators.
Backtesting and optimization
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Backtest this Pine strategy on the current daily chart with ₹1,00,000 capital and 5 bps commission." | tl-backtesting | backtesting.capabilities, backtesting.run (language pine), backtesting.get | Metrics, trades and equity, after polling to completion |
| "Run my saved Python strategy in sequential mode on the last 250 candles." | tl-backtesting | backtesting.run, backtesting.get | A sequential run (capped at 250 candles) |
| "Compare my last two runs and explain the difference." | tl-backtesting | backtesting.list, backtesting.compare | A side-by-side comparison, or an explicit incompatibility error |
| "Save this run to my library and export the trades." | tl-backtesting | backtesting.librarySave, backtesting.export | An archived snapshot and exported JSON |
| "Preview a walk-forward test of fast 10–20 and slow 40–60 with 3 folds." | tl-strategy-optimization | backtesting.optimizePreview | Candidate grid, fold dates and workload, with no execution |
| "Run that walk-forward optimization and report out-of-sample results." | tl-strategy-optimization | backtesting.optimize, backtesting.get (optimization section) | Per-fold training winners and their held-out test results |
| "Rerun the selected parameters as a new backtest and make an HTML report." | tl-strategy-optimization | backtesting.optimizeRerun, backtesting.optimizeReport | A new run plus a detailed offline report |
Grids are limited to 4 parameters and 12 combinations, with 1–5 walk-forward folds. Read optimize without overfitting and the backtest with an AI assistant walkthrough.
Bar replay and paper trading
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Start bar replay on this chart; I'll pick the start candle." | tl-charts | chart.replayControl | Replay mode ready; you choose the candle |
| "From the paused replay cursor, start a paper session with ₹50,000." | tl-paper-trading | backtesting.practice, backtesting.get | A virtual-money session at the cursor |
| "Buy 10 shares at market, add a stop at 2% below, then step 5 candles." | tl-paper-trading | backtesting.order, backtesting.protect, backtesting.step | Simulated fills and protection, read back |
| "Add HDFCBANK to this practice session." | tl-paper-trading | backtesting.addInstrument | A second instrument sharing the same balance (up to 10) |
| "Forward-test my completed backtest on live candles as paper." | tl-paper-trading | backtesting.forward | A forward paper session; no broker orders |
See bar replay, paper trading and practising with replay.
Historical Patterns and research pages
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "For NIFTY, study Monday 09:15–10:15 IST candles over the last year." | tl-patterns | patterns.show, patterns.configure, patterns.run, patterns.getResults | Coverage and filtered statistics for that window |
| "Open the worst occurrence on the chart." | tl-patterns | patterns.openOccurrence | The chart navigated to that dated window |
| "Publish a one-page HTML summary of my last backtest and show it." | tl-research | backtesting.get, research.create, research.show | A saved report in the right sidebar (no scripts, max 50 reports) |
Historical Patterns is a time-window study, not chart-pattern detection; see Historical Patterns. Saved HTML pages are described on research reports.
Demo recording and testing
| You ask | Skill | Main commands | You get back |
|---|---|---|---|
| "Record a 30-second 60 FPS clip of you adding RSI to this chart." | tl-desktop | record-start, ui-click, record-stop, record-status | A private local video file path |
| "Record what I do next and turn it into a replayable script." | tl-desktop | ui-codegen-start, ui-codegen-stop, ui-run | A ui-run step batch; password and payment fields are left out |
| "Open our pricing page in the demo browser and highlight 'Pro'." | tl-desktop | demo-browser-open, ui-highlight-text | A separate-session demo window with a highlighted phrase |
| "Run the smoke test suite and report failures." | tl-testing | Suite steps built from catalog commands | A pass/fail report |
Details: AI demo recording.
What not to ask
- Live orders, broker login or "connect my Zerodha". These are human-only or not available at all.
- Automatic chart-pattern detection. Historical Patterns is a time-window study, not a pattern finder.
- "Find the best parameters." The optimizer reports in-sample and held-out evidence. Ask for walk-forward results, not a winner.
- Natural-language scans with no rules. The assistant must translate your idea into the screener's supported fields.
Steps
- Install TradingLead Desktop for Apple Silicon Mac and enable Help > AI integrations
- Open a new terminal and start your assistant
- Ask it to run tradinglead agent-guide and target your active chart tab
- Copy a request from this cookbook and adjust the symbol, interval and numbers
- Check the returned evidence and the visible chart before acting on anything
Limitations
- Needs TradingLead Desktop with managed AI integration (Apple Silicon Mac 0.1.8 or later; the public Windows 0.1.7 build lacks it).
- Commands available depend on the open tab's live catalog and the deployed website; ask the assistant to check capabilities first.
- No live broker orders, broker sign-in or permission approval by the assistant.
- Backtests, optimizations and scans are research on historical data, not advice or proof of future returns.
- TradingLead does not provide or bill for the AI model.
Frequently asked questions
- Which AI assistants can use these prompts?
- Any assistant that can run terminal commands, such as GitHub Copilot CLI or Claude Code, through TradingLead Desktop. GitHub Copilot gets the TradingLead skills registered automatically; other agents read them with tradinglead agent-guide and tradinglead skills read.
- Can I ask the assistant to buy or sell a stock for me?
- No. TradingLead has no live broker order commands. Requests about orders work only as simulated paper orders inside a replay or forward paper session.
- Can the AI find chart patterns like head and shoulders automatically?
- No. The tl-patterns skill runs Historical Patterns, which studies one ticker across weekdays and IST time windows. Automatic chart-pattern detection is not a TradingLead feature.
- Why does the assistant ask which tab to use?
- The skills require an explicit tab ID for every command, even with one open tab, so changes never land on the wrong chart or account.
- Do these prompts use my scan or data quota?
- Some do. Scans, Historical Patterns runs, chart reloads, history requests and backtests can use account or provider allowance. Reads such as watchlists.getRows or chart.getState do not.
Related
Last updated: