메뉴
Daily XAUUSD gold trading signal for rakyat marhaen. One-page PDF, dark OANDA-style chart, zoomed candlesticks, buy/sell zones near current price, EMA 20/50, R:R visual, simple language. Not broker jargon.
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 | daily-trading-signal-briefing |
| description | Daily XAUUSD gold trading signal for rakyat marhaen. One-page PDF, dark OANDA-style chart, zoomed candlesticks, buy/sell zones near current price, EMA 20/50, R:R visual, simple language. Not broker jargon. |
| triggers | ["gold signal","trading signal","XAUUSD signal","daily gold","emas signal"] |
| tags | ["trading","xauusd","gold","signal","daily","pdf"] |
Rakyat marhaen — common people, not brokers. Simple language. BUY HERE, SELL HERE, SL HERE. No Fibonacci, no Elliott Wave jargon.
Abang Udin (@rico_ricaldo_33) — pure technical trader, OANDA platform, XAUUSD only.
LONG ONLY. Short is not in the playbook. R:R on short near oversold support = terrible. Short only becomes relevant if price breaks major support with confirmation.
Two signals every morning:
Jiwa belum kuat = NORMAL. This is preparation. Watch first. Learn the zones. Paper trade mentally. When jiwa ready, the signal will be there. No pressure. Market sentiasa ada esok.
Preferred renderer: /root/trading/scripts/chart_pro.py — matplotlib-based, 180 DPI, dark theme, mobile-ready. Replaces older chart approaches.
cd /root/trading && python3 scripts/chart_pro.py --signal SHORT --entry 4055 --sl 4076 --tp 4027 --json
Output: /tmp/xauusd_chart.png (300KB, 180 DPI, landscape)
JSON metadata: price, bias, confidence, rsi, ema20/50/200, support[], resistance[], signal, entry, sl, tp
Features: candlesticks, EMA 20/50/200, S/R levels (max 2-3 clean), signal overlay with zone shading, RSI panel, bias pill, confidence %, bottom synthesis banner.
$ in matplotlib text = LaTeX math mode crash. Dollar signs in any text passed to ax.text(), fig.text(), ax.set_title() trigger ParseException: Expected end of text, found '$'. Fix: replace $ with USD in all matplotlib text strings, OR set plt.rcParams['text.usetex'] = False + plt.rcParams['mathtext.default'] = 'regular' before any plot calls. The chart_pro.py already handles this.--json, numpy types (int64, float64) crash json.dumps() with TypeError: Object of type int64 is not JSON serializable. Fix: use a custom JSON encoder that converts np.integer → int, np.floating → float, np.ndarray → list. Already patched in chart_pro.py (2026-07-16). General rule: ANY script that returns JSON from numpy/pandas data MUST use a custom encoder — never assume json.dumps() handles numpy types.execute_code() sandbox lacks matplotlib. The hermes_tools sandbox uses a different venv. Use write_file() + terminal() for chart generation. chart_pro.py runs via terminal.python3 in cron context, the shell wrapper may try to source arifOS auto-venv functions (_arifos_auto_venv, _arifos_auto_env) which fail with command not found. Always use /root/venv/bin/python3 explicitly for cron-triggered chart/engine scripts, not bare python3.If chart_pro.py fails, fall back to the template at templates/gold_signal_chart.py (also matplotlib, same dark theme, slightly different layout).
When /root/trading/ is entirely missing, use the yfinance OHLC fallback chain documented in references/chart-fallback-yfinance.md (proven 2026-07-20).
/root/GOLD-SIGNAL-{DATE}.pdf| LEVEL | |
|---|---|
| ENTRY | $X - $Y |
| STOP LOSS | $Z |
| TARGET 1 | $T1 (+$N) |
| TARGET 2 | $T2 (+$N) |
| R:R | 1:X.X / 1:Y.Y |
| STATUS | READY / WAIT FOR DATA / NO TRADE |
| LEVEL | ACTION | |
|---|---|---|
| TAKE PROFIT 1 | $TP1 | Jual separuh (50%) |
| TAKE PROFIT 2 | $TP2 | Jual semua (100%) |
| TRAIL STOP | $X | Naikkan SL ke entry |
| EMERGENCY EXIT | $EX | Cut loss — jangan tunggu |
web_search("XAUUSD gold price today [DATE]")
web_search("gold news today [MONTH] [YEAR] economic events")
web_search("XAUUSD technical analysis support resistance [MONTH] [YEAR]")
Then web_extract() the best technical analysis page (DailyForex, Vantage, FXStreet) for concrete levels.
Key data to extract:
Use extracted support/resistance as zone boundaries — NOT formulaic offsets. Buy zone = nearest strong support cluster (±$10). Sell zone = nearest strong resistance cluster (±$15). SL = below/above the zone by $15-25 (tight, not wide). Targets = next resistance (for buys) or next support (for sells).
Write chart Python to /tmp/build_gold_signal.py, run via terminal().
IMPORTANT: Use write_file() + terminal(), NOT execute_code().
Matplotlib/reportlab are installed in the system Python but execute_code runs in a sandbox that may not have them.
Write PDF Python to /tmp/build_gold_pdf.py, run via terminal().
Uses reportlab — same pattern: write_file() + terminal().
CRITICAL: Copy ALL signal-level variables (BUY_ZONE_LOW, BUY_ZONE_HIGH, SL, T1, T2, ENTRY_MID) into the PDF script header. The PDF runs in a separate process — it does NOT inherit variables from the chart script. Use f-strings in the data tables so values update automatically when variables change.
Send chart as photo + PDF as document to the user. Include strategy table in message text for quick reading.
references/gold-api-extension-pattern.md — how to add new endpoints to gold-api (THE way to add trading features)references/red-news-impact.md — CPI/NFP/FOMC impact on gold: move sizes, 15-minute rule, Syed briefing textreferences/gold-data-apis.md — API endpoints and pricing data sourcesreferences/proven-data-pipeline.md — proven web-scraping pipeline (no API key needed), pitfalls, source quality rankingreferences/ohlc-estimation-method.md — how to estimate daily OHLC from web search when Twelve Data is unavailable (proven 2026-07-15)templates/gold_signal_chart.py — OANDA dark candlestick chart (matplotlib, proven 2026-07-14)templates/gold_signal_pdf.py — one-page PDF with strategy table (reportlab, proven 2026-07-14)For interactive live charts, deploy TradingView lightweight-charts web app:
gold-api/var/www/html/gold/index.htmltrading-signal-chart skill → references/tradingview-lightweight-charts-webapp.mdUse web chart for live monitoring. Use PDF for daily briefings and signal delivery.
See: references/gold-dashboard-architecture.md — full architecture, Caddy routing, trailing-slash bug (fixed 2026-07-17), stale data investigation, zone computation logic, API endpoints, troubleshooting commands, and how to extend gold-api.
The gold-api (port 3456) now has endpoints that replace direct yfinance calls:
# APEX market intelligence (G/C_dark/dS/state/verdict)
curl -sf localhost:3456/api/gold/apex
# Full signal from engine_v2 + position sizing
curl -sf localhost:3456/api/gold/signal_v2
# ForexFactory economic calendar (high-impact USD events)
curl -sf localhost:3456/api/gold/calendar
# Quick ticker (price, change, RSI, EMAs)
curl -sf localhost:3456/api/gold/ticker
When extending gold-api (adding new endpoints):
/var/www/html/gold/api/fetch_gold.py (function cmd_XXX)fetch_gold.py/var/www/html/gold/api/server.js handlers object/gold/ prefix) for Caddy strip_prefix compatibilitysystemctl restart gold-apicurl -sf localhost:3456/api/XXXCRITICAL USER RULE: Arif forbids creating new servers/repos. Always extend the existing gold-api. Never create a new MCP server, new port, or new service for trading features.
Command dependencies: Not all endpoints need the /root/trading module. ticker, history, signals, levels, macro, calendar are self-contained. signal_v2 and apex require the trading package. If those endpoints return HTTP 500, see references/gold-api-command-dependencies.md for the degraded-mode fallback.
Human-readable labels: All user-facing text (gold site, Telegram messages to Syed) must use plain language. NO jargon like G, C_dark, dS, APEX, Φ. Use: Clarity, Risk, Trend Energy, Condition, Structure, Strength, Signal, Stability, Agreement, Momentum, Volume. State names: Clear Trend / Ranging / Choppy. Verdicts: Strong Signal / Good to Go / Wait / Hold Off.
--briefing has a session filter that blocks output during Asian session (00:00-07:00 UTC). The 8am MYT cron (00:00 UTC) ALWAYS hits this gate. When the engine returns only "SESSION: ASIAN/BLOCKED — Outside trading hours" with no signal data, fall back to the web-search data pipeline (Step 1-2). Do NOT deliver just the session-blocked message — the user needs actual signal levels. The engine computes the signal internally but format_briefing() returns early when session_ok=False. To bypass for testing, edit gold_engine.py line 343-346 to remove the early return, or run during London/NY hours.entry_mid = (BUY_ZONE_LOW + BUY_ZONE_HIGH) / 2. The PDF must use the SAME entry midpoint, not a different hardcoded value. If chart says 1:1.6 / 1:2.5, PDF must say the same. Derive PDF values from the chart's entry_mid: risk = entry_mid - SL, reward1 = T1 - entry_mid, reward2 = T2 - entry_mid./tmp/build_gold_signal.py and /tmp/build_gold_pdf.py). They run in separate processes — no shared state. The PDF script MUST independently define ALL signal-level variables: BUY_ZONE_LOW, BUY_ZONE_HIGH, SL, T1, T2, ENTRY_MID. Common mistake: defining ENTRY_MID, SL, T1, T2 in the PDF script but forgetting BUY_ZONE_LOW/BUY_ZONE_HIGH (used in the verdict line and zone labels). Fix: copy the full signal-level variable block from the chart script into the PDF script header.support[] and resistance[] arrays in the JSON output. If levels are within $5 of each other, treat them as ONE level and supplement with web-search S/R data from Step 1. The engine's S/R detection works best on volatile days with clear swing highs/lows — on quiet/consolidation days, the algorithm clusters.execute_code() for matplotlib/reportlab — use write_file() + terminal() (packages are in system Python, not in sandbox)/root/venv/bin/python3, not bare python3. The shell wrapper in cron context tries to source arifOS auto-venv functions that don't exist, causing command not found. Affects: price_alert.py, gold_engine.py, chart_pro.py, any script run via terminal() in cron jobs.position_sizer.py calls cfg.contract_multiplier but TradingConfig didn't have this field. Fix: added contract_multiplier: float = 1.0 to /root/trading/core/config.py. If other config attributes fail with AttributeError, add them to the dataclass.cmd_signal_v2 and cmd_apex in fetch_gold.py import from trading.* (lines 420-424, 515-521). If /root/trading/ doesn't exist on the server, these endpoints return HTTP 500 with "No module named 'trading'". The cmd_ticker function is SELF-CONTAINED (only uses yfinance + numpy/pandas directly) and works regardless. Degraded-mode fallback for cron jobs: When curl -sf localhost:3456/api/gold/signal_v2 returns exit code 22 (HTTP error), check systemctl status gold-api --no-pager -l for the error. If it's "No module named 'trading'", fall back to ticker-only mode: use curl -sf localhost:3456/api/gold/ticker for price/RSI/EMAs/S/R levels, supplement with web search (hound mcp_smart_search + mcp_smart_fetch) for narrative context and concrete levels from FXEmpire/FXStreet. The ticker's built-in signal field and S/R arrays are sufficient for the decision gate. Chart/PDF generation still works if /root/trading has chart_pro.py — but check existence first with test -f /root/trading/scripts/chart_pro.py.A Python-based signal engine is live at /root/trading/scripts/gold_engine.py. It can replace or supplement the web-search-based signal generation above.
Use the engine when: you need fast, consistent, automated signals with proper indicator calculation AND it's during London/NY hours (15:00-05:00 MYT). Use the web-search method when: you need narrative context, news sentiment, the engine data source fails, OR the engine returns session-blocked (which happens at 8am MYT cron — see Pitfalls).
cd /root/trading && python3 scripts/gold_engine.py --briefing
Engine workflow for cron jobs (8am MYT): Try engine first. If output is ≤3 lines and contains "ASIAN/BLOCKED", discard it and fall back to the web-search pipeline (Step 1-2 above). Never deliver just the session-blocked message.
Engine features: EMA 20/50, RSI + divergence, candle patterns, S/R levels, DXY/yields macro, session filter (London/NY only), ≥2 confluence rule, journal logging.
Cron job 2258f1b3fa0e delivers daily at 8am MYT to SADO group.
Confluence rule (F3 WITNESS): Single-indicator signal = rejected. Minimum 2 of: EMA crossover, S/R proximity, candle pattern, RSI divergence.
Cron job 2258f1b3fa0e adds a voice note (BM OsmanNeural, ~90 seconds) alongside the text + chart briefing for the SADO group. This audio channel reinforces the same signal in three formats (text + chart + voice) — proven pattern from tts-edge-fallback and syedos skills.
Voice generation step (added after Step 5 in the cron prompt):
After text message is formatted, populate /tmp/syed_voice.txt from the same data:
Numbers — spelled out as BM words. Examples:
Generate MP3:
cat > /tmp/syed_voice.txt << 'EOF'
[populated script]
EOF
edge-tts --voice ms-MY-OsmanNeural --rate "+5%" \
--file /tmp/syed_voice.txt --write-media /tmp/syed_voice.mp3
Deliver: text first → chart → voice note (voice goes last in Telegram). All three to SADO group.
Voice-only skip rule: If verdict is SABAR AND state is "Choppy", voice is optional (text suffices — "jangan trade" is the only message). Otherwise, voice is MANDATORY.
Full template + verdict→action mapping + edge-tts gotchas: see references/voice-briefing-format.md in syedos skill.
Three monster events move gold $50-150 in minutes:
| Event | MYT Time | Avg Gold Move | Direction Rule |
|---|---|---|---|
| CPI | 20:30 monthly | $40-80 (1-2%) | CPI low → Gold up. CPI high → Gold down |
| NFP | 20:30 first Friday | $50-100 (1.2-2.5%) | Jobs few → Gold up. Jobs many → Gold down |
| FOMC | 02:00, 8x/year | $30-80 + $50-100 press | Dovish → Gold up. Hawkish → Gold down |
Why? Spread blows out 100-200 pips. Stop hunt. Slippage. First move is always wrong.
"Abang Sado, setiap kali ada red news — CPI, NFP, FOMC — tutup semua position 15 minit sebelum. Tunggu 30 minit lepas news. Baru masuk balik. Lot kurangkan 50%. Spread masa red news boleh makan semua profit micro-lot kau."
Full reference: references/red-news-impact.md
Arif explicitly forbids APEX theory jargon in gold site and Telegram messages to Syed.
| Internal Term | Human Label |
|---|---|
| G score | Clarity |
| C_dark | Risk |
| dS/dt | Trend Energy |
| A | Structure |
| P | Strength |
| E | Signal |
| X | Stability |
| Φ (Phi) | Agreement |
| CLARITY state | Clear Trend |
| STABLE state | Ranging |
| CHAOS state | Choppy |
| BUY direction | Trending Up |
| SELL direction | Trending Down |
| FLAT | No Direction |
| SEAL verdict | Strong Signal |
| PROCEED | Good to Go |
| SABAR | Wait |
| HOLD | Hold Off |
NEVER use G, C_dark, dS, APEX, Φ in user-facing text. Always use the human label.
Reconcile split-brain truth sources in a multi-organ federation — when arifOS capability_map, Hermes config.yaml providers block, MCP server registry, and runtime env all claim to be the canonical source for "what's configured". Load when Arif reports "Hermes says 1 provider, arifOS says 28", when a tool call fails with unknown provider, or when fixing audit findings about API key / model / organ coverage.
Add, fix, or migrate custom/third-party LLM providers in a Hermes Agent install so they appear in `/model` (CLI) and the Telegram gateway picker. Covers providers block shape, env-var key mapping, security-blocker on direct config writes, three-tier multimodal fallback chains, MiMo UltraSpeed/Pro/v2.5 capability matrix and supports_vision flag for forced auxiliary routing, vault→hermes.env key sync pattern, picker zen with 5-state failure classification (OK/PAUSED/BROKEN/CHAT_INCOMPATIBLE/EXC) and dedupe by route priority, MoA preset cross-dependency on provider models, MiniMax direct provider + mmx-cli for multimodal generation, OpenCode Go vs Zen tier verification, and dead fallback cleanup.
MiniMax multimodal via mmx-cli — TTS, video, music, image, vision, search
Test whether a proposed improvement actually beats the existing baseline BEFORE committing to build it. Spawn parallel forge agents that each compare challenger vs baseline on the same data. Kill ideas that don't improve; forge only what measurably helps.
Map of the arifOS federation — organs, ports, roles, MCP surfaces, health probes. Load when Arif asks 'what organs are alive', 'show me the federation', 'is X running', 'what port is Y', or when you need to route a task to the right organ.
Probe live state BEFORE proposing fixes, additions, or changes. Read configs, test commands, check what exists — then act only on real gaps. Use ALWAYS. This is not optional. Arif corrected multiple times (2026-07-04, 2026-07-08, 2026-07-11) for proposing fixes to things that already worked, for citing stale audits as live state, and for accepting 'first time' / 'novel' / 'pick from N options' prompts without probing disk.