crypto-chart-analysis

Crypto Chart Analysis (Cloud Agent)

Educational market-structure framework for reading crypto charts from a Cursor Cloud Agent VM. Not investment advice. Never promise profits, “sure” entries, or guaranteed targets.

When to use

• User shares a ticker, chart screenshot, timeframe, or setup (long/short)
• Asks for structure, levels, bias, invalidation, R-multiple
• Asks about OI, funding, liquidations, breakouts, or alt vs BTC context
• Cloud session: “разбери BTC”, “что по ETH”, “оценка сетапа”

When not to use

• Non-market coding or infra work
• Guaranteed signals / “what to buy now” as personalized advice
• Pure news/tokenomics without a chart ask

Priority weights (keep this order)

Weight	Layer	Job
~70%	Structure + risk	Regime, levels, MTF, invalidation, size logic
~20%	Volume / flow / positioning	Confirm or challenge the structure story
~10%	Oscillators / named patterns	Timing only, never the thesis

Hard rule: structure → risk/invalidation → volume/flow/OI-funding → indicators. Reverse order is a failure mode.

---

Cloud Agent runtime (mandatory here)

This skill is meant to run in the current Cloud Agent environment, not on the user’s phone or local TradingView.

Skill location

Resolve the skill directory in this order (first hit wins):

1. $CRYPTO_CHART_SKILL_DIR if set
2. /cursor/stores/self/skills/crypto-chart-analysis
3. /agent/skills/crypto-chart-analysis
4. .cursor/skills/crypto-chart-analysis under the workspace (if a repo is mounted)

Fetcher script: <skill_dir>/fetch_market_context.py

Before analysis — pull live data

Unless the user only wants theory, always fetch before concluding:

python3 "<skill_dir>/fetch_market_context.py" \
  --symbol BTCUSDT \
  --market perp \
  --venue auto \
  --tfs 1d,4h,1h \
  --limit 120 \
  -o /tmp/crypto_ctx.json

Then Read /tmp/crypto_ctx.json (or use stdout). Defaults:

Flag	Default	Notes
--market	perp	spot for cash markets
--venue auto	okx (perp) / binance_vision (spot)	see egress table
--tfs	1d,4h,1h	add 15m only for timing
--limit	120	enough for swings + ATR

Do not pass --full-candles into the model context unless needed — default payload keeps candles_tail + swings.

Egress reality (this Cloud VM)

Source	Typical status	Use for
OKX (www.okx.com)	OK	Perp klines, funding, OI — default
Bitget	OK	Perp fallback
Binance Vision (data-api.binance.vision)	OK	Spot klines
Binance main / fapi	Often geo-blocked	Avoid as primary
Bybit	Often geo-blocked	Avoid as primary
CoinGecko	OK	Quick spot price sanity check only

If a venue fails, retry --venue okx or --venue bitget. State which venue’s book you used — never pretend multi-exchange aggregation.

Screenshots

If the user attaches a chart image: Read the image path, extract marked levels/TF, still fetch live OHLCV/OI/funding to timestamp the analysis. Call out conflicts between the screenshot and live data.

Tools to use (cloud)

• Shell — run fetch_market_context.py / curl fallbacks
• Read — JSON context + chart images
• WebSearch / WebFetch — only for scheduled events (CPI, options expiry), not as a candle source
• WriteCanvas — optional durable bias map for long analyses
• Do not claim you opened TradingView or controlled the user’s phone

Cloud limits (be explicit)

• No true DOM/tape/footprint from this skill — treat order flow as unavailable unless user pastes it
• OI/funding are venue-specific (OKX ≠ Binance)
• Script structure_hint is a helper, not the final call — agent still applies the full checklist
• No private account balances, positions, or API keys required (public market data only)

---

Workflow

0. Intake + fetch

Collect if missing: symbol, spot vs perp, TFs, goal (bias / entry sketch / critique).

Then run the fetcher. If data incomplete, produce a conditional analysis with assumptions.

1. Regime (primary)

• Trend up (HH/HL), trend down (LH/LL), or range
• Last structure shift in plain language
• HTF owns bias; LTF for timing — do not start from 1m

Use script swings + structure_hint as input, then judge.

2. Liquidity and S/R (primary)

Zones, not 1-tick lines: swing H/L, range edges, round numbers, stop pools above highs / below lows. Label support / resistance / flip.

3. Risk first (primary)

Every actionable idea needs invalidation, stop logic, R-targets, and explicit NO TRADE when mid-range / TF conflict / illiquid / pre-event fog. No invalidation → no direction.

4–5. Volume + volatility

Participation vs empty breakouts; compression vs expansion (use atr14 from fetch). Spot vol ≠ perp vol.

6. Crypto positioning (from fetch positioning)

Reading	Hypothesis (not certainty)
Price ↑ + OI ↑	Fresh leverage
Price ↑ + OI ↓	Short cover
Price ↓ + OI ↑	Fresh shorts
Price ↓ + OI ↓	Long washout
Extreme funding	Crowding — not auto-reversal
Liquidations	Event/fuel — need acceptance after

Use funding_apr_pct_approx from the script. Pair funding with OI + structure always.

7. Secondary tools (timing only)

Candles / RSI / MA / Fib / pattern names only at zones after structure+risk. Skip holy-grail indicator stacks.

8. Practitioner checks

No-trade zones; breakout = hypothesis until acceptance; liquidity above highs; scenario + anti-scenario; session/events; BTC beta; fees/funding/slippage; journalable one-liner.

9. Output format

## Bias map
- Instrument / venue / TFs / data timestamp (UTC):
- Regime (HTF → LTF):
- Key zones:

## Working hypothesis
- Idea (or NO TRADE) + why (structure first)
- Confirm / invalidate

## Risk sketch
- Invalidation / R / no-trade note

## Positioning (perp)
- OI + funding APR (venue-named)

## Secondary timing (optional)

## Anti-scenario

## Assumptions / missing data
- e.g. no liquidation feed, single-venue OI

One-line disclaimer: framework, not financial advice.

---

Response style

• Match user language (RU/EN)
• Conditional (“if held / if broken”); prefer NO TRADE over forced bias
• No hype, no emojis, no guaranteed targets
• Cite venue + fetch timestamp

Anti-patterns

• Analyze from memory without a fetch when live price matters
• Lead with RSI/MACD before structure
• Treat funding extremes or liquidations as automatic reversals
• Entries without invalidation
• Claiming Binance OI while data came from OKX
• Presenting script hints as certified signals

Files in this skill

File	Role
SKILL.md	This playbook
fetch_market_context.py	Cloud data pull + swings/ATR/funding APR
reference.md	Deeper definitions / matrices

If the user wants more depth on a subsection, open reference.md.