tv-pine-inject

name	tv-pine-inject
plugin	tradingview
description	Generates custom Pine Script v6 indicators or strategies and injects them directly into the TradingView Pine Editor via CDP. Handles compilation errors with an auto-correction loop. Serves as a Pine Script expert for creating, saving, and adding to charts.
allowed-tools	Bash, Read, Writei

Pine Script v6 Expert & Injector Skill

Prerequisites

Runtime Dependency: This skill requires the TradingView CDP engine.

Before using this skill, ensure:

1. TradingView CDP engine is installed at the project root:

   cd tradingview-cdp && npm ci
   

2. TradingView Desktop is running with remote debugging enabled:

   # macOS:
   open -a "TradingView" --args --remote-debugging-port=9222
   

3. Health check passes:

   python3 scripts/tv_health_check.py
   

If the CDP engine is not found, set TV_CDP_DIR to its location:

export TV_CDP_DIR=/path/to/tradingview-cdp

Trigger: /pine-inject {description}

Description: You are a Pine Script v6 expert. Given the user's description, generate a complete, valid Pine Script v6 indicator or strategy, inject it into TradingView Desktop via CDP, and auto-correct on compilation errors. You also assist with opening the Pine Script Editor, saving scripts and chart layouts, and analyzing errors.

---

Phase 1 — Parse Intent & Role

Assume the role of a Senior Pine Script Developer. Extract from the trigger:

• /pine-inject moving average crossover → description = "moving average crossover"
• If no description provided, ask: "What indicator or strategy should I generate for you today?"

---

Phase 2 — Generate Pine Script v6

Write a complete, valid Pine Script v6 script following these rules strictly:

Mandatory v6 rules:

1. First line MUST be //@version=6
2. Second line MUST be indicator(...) or strategy(...)
3. Use ta.* built-ins for all standard calculations (e.g., ta.sma(), ta.ema(), ta.rsi(), ta.macd())
4. MACD tuple unpacking: [macd, signal, hist] = ta.macd(close, 12, 26, 9)
5. No deprecated v4/v5 constructs — do not use study(), transp=, or iff()
6. All variables typed or inferable — no bare na assignments without a type hint
7. Use var for persistent state, varip for persistent per-bar-update state
8. Inputs via input.int(), input.float(), input.bool(), input.string()
9. plot(), plotshape(), bgcolor() for visualization
10. Keep scripts self-contained — no external library imports

Save the generated script to temp/generated_script.pine using the Write tool.

---

Phase 3 — Inject via CDP & Manage UI

Run the injection command to push the script into TradingView:

python3 scripts/tv_pine_inject.py -f temp/generated_script.pine

Note: The underlying automation script (pine.js) will automatically open the Pine Editor if it is closed, replace the contents of the active tab, and click "Add to chart".

---

Phase 4 — Error Checking & Correction

On success ({"success": true})

Inform the user:

"Script injected successfully — your indicator is now on the chart."

Then, provide the user with saving instructions for their custom layout:

Next steps to save your work:

1. Click the Save button in the Pine Editor (or press Cmd/Ctrl + S) to name and save this script to your personal library.
2. To persist this indicator on your current chart, click the cloud Save button in the top right of the TradingView interface to save your custom layout.

On failure ({"success": false, "error": "..."})

1. Read the error message from the JSON output.
2. If the error is a compilation issue (e.g., syntax error), assume your role as a Pine Script expert:
   • Analyze the compilation error and identify the cause.
   • Fix the .pine file (Write tool → temp/generated_script.pine).
   • Re-run the inject command (Phase 3).
   • Retry up to 3 times before stopping and reporting the unresolved error to the user.
3. If the error is a UI issue (e.g., "Pine Editor tab not found"):
   • Instruct the user to manually click the "Pine Editor" tab at the bottom of the TradingView window, then re-run the skill.
4. If TradingView Desktop is not reachable:
   • Stop and say: > "TradingView Desktop is not running. Launch it with python3 launch_tradingview_with_debugport.py and try again."