Skill Runtime Lookup Guide

Audience: a MobiusQuant Skill running inside an AI Agent tool (Claude Code, Codex, OpenClaw, Hermes, or similar). This page tells the Skill where to fetch what, in what order, when the end user asks about indicators or market symbols.
Paste-friendly: drop a link to this page into your SKILL.md, or paste the relevant sections inline.

TL;DR

User intent	Best source	Why
"What indicators do you have?"	`GET /llms-full.txt` (this site)	Read the complete English reference in one request
"What params does `X` take?"	`GET /indicators/<id>.md` (this site)	Structured English, no narrative noise
"How do I interpret `X` signals?"	`GET /zh/indicators/<id>.md` (this site)	Chinese narrative — interpretation, caveats, focus points
"What's the canonical symbol for `<alias>`?"	`GET /api/symbols/search?q=<alias>` (engine)	Live fuzzy lookup, includes Chinese/English aliases
"Is `<symbol>` on `<venue>`?"	`GET /api/markets/<exchange>/<market>/symbols` (engine)	Authoritative live list
"What's BTC's 5m kline right now?"	`GET /api/klines` or `POST /api/indicators` (engine)	Live data — always use the API, never cite cached docs
"How do I interpret these indicator values?"	`POST /api/indicators?explain=true` (engine)	Response carries `explain.summary_focus` / `signals` / `caveats` — pass these to your LLM verbatim, do NOT have it guess interpretation rules

Rule of thumb: static reference info → fetch from docs.mobiusquant.ai. Live data → call api.mobiusquant.ai. Indicator interpretation rubrics → always include ?explain=true and forward the explain field to your LLM.

Static reference (docs.mobiusquant.ai)

This site is public, CORS-enabled, no token required. Fetch and cache freely.

Bootstrapping: one-shot full reference

http

GET https://docs.mobiusquant.ai/llms-full.txt

Returns structured English covering API token rules, 13 supported venues including gold spot from 5m, all indicators (with default params + output columns), Skill integration, and AI agent guidance.

Recommended: fetch once at Skill startup (or first time the user mentions data/indicators). Cache for the session.

Per-indicator detail

http

GET https://docs.mobiusquant.ai/indicators/<id>.md         # English (structured)
GET https://docs.mobiusquant.ai/zh/indicators/<id>.md      # Chinese (with interpretation)

Where <id> is the canonical lowercase ID. Indicator IDs:

accelerator_oscillator awesome_oscillator cci chande_mo connors_rsi
coppock_curve dpo kst macd momentum ppo roc smi_ergodic stoch_rsi
trix tsi ultimate_oscillator williams_r          # momentum (18)

fisher_transform rci_ribbon rsi rvi              # oscillator (4)

alligator alma ema hma hmm_regime lsma mcginley_dynamic
mobius_trend parabolic_sar supertrend tema
pi_cycle_bottom smc vwma williams_stops          # trend (15)

adx aroon_oscillator choppiness_index vortex     # trend_strength (4)

atr bollinger chande_kroll_stop donchian_channels
fibonacci_bollinger_bands historical_volatility
keltner_channels linear_regression_channel
mass_index stdev                                 # volatility (10)

chaikin_oscillator cmf cvd kvo obv pvo vwap      # volume (7)

price                                            # base (1)

Möbius Lab (proprietary): mobius_trend. These are MobiusQuant in-house indicators, not classic public ones.

URL	Use
`/llms.txt`	Site index (English) — overview before deep-diving
`/llms-full.txt`	Full English reference — recommended first fetch
`/indicators/index.md`	English catalog overview
`/zh/indicators/index.md`	Chinese catalog overview
`/markets.md`	English venue + canonical-symbol rules
`/zh/markets.md`	Chinese venue + enumerated symbol lists, including gold spot from 5m
`/token.md`	Token application + Bearer auth + rate limits
`/agents.md`	General AI Agent operating manual

All HTML pages also resolve as .md at the same path. CORS * on every .md and .txt.

Live data (api.mobiusquant.ai)

Optional Authorization: Bearer mq_<43chars>. Without a token you still get 10 req/min per IP (good for exploration). Add a token to upgrade to 60 req/min. Get one free at https://www.mobiusquant.ai/zh/apply-token (7-day anonymous) or https://www.mobiusquant.ai/zh/account (logged-in permanent).

Rate limit: anonymous 10 req/min per IP · with token 60 req/min.

Discovery endpoints (live, authoritative)

bash

# All venues + WS health
curl -H "Authorization: Bearer mq_xxx" https://api.mobiusquant.ai/api/markets

# Symbols on one venue
curl -H "Authorization: Bearer mq_xxx" \
     https://api.mobiusquant.ai/api/markets/binance/perp/symbols

# Live indicator catalog (params + output cols)
curl -H "Authorization: Bearer mq_xxx" https://api.mobiusquant.ai/api/indicators/registry

# Chartable subset + drawing protocol
curl -H "Authorization: Bearer mq_xxx" https://api.mobiusquant.ai/api/indicators/catalog

# Interval support matrix
curl -H "Authorization: Bearer mq_xxx" https://api.mobiusquant.ai/api/intervals

Natural-language symbol resolution

The user says "BTC", "比特币", "Tencent", "茅台" — let the engine resolve:

bash

curl -H "Authorization: Bearer mq_xxx" \
     "https://api.mobiusquant.ai/api/symbols/search?q=比特币"
# → [{"canonical": "BTCUSDT", "exchange": "binance", "market": "perp", ...}]

Or fetch the full mapping table once:

bash

curl -H "Authorization: Bearer mq_xxx" https://api.mobiusquant.ai/api/symbols/builtin

Compute / data endpoints

bash

# Raw OHLCV
GET https://api.mobiusquant.ai/api/klines?exchange=binance&market=perp&symbol=BTCUSDT&interval=5m&limit=100

# Indicators on klines
POST https://api.mobiusquant.ai/api/indicators
  body: {
    "exchange":"binance","market":"perp","symbol":"BTCUSDT",
    "interval":"5m","limit":100,
    "calc":[{"name":"rsi","params":{"period":14}}]
  }

# Same + chart drawing protocol
POST https://api.mobiusquant.ai/api/chart

# Compute indicators on caller-supplied OHLCV (backtest / custom feed)
POST https://api.mobiusquant.ai/api/indicators/compute
  body: {
    "klines": [[1735689600000, 95000, 95500, 94800, 95200, 1234.5], ...],
    "interval": "1h",
    "calc": [{"name":"rsi","params":{"period":14}}]
  }
  # ≤ 2000 bars; HMM-only requests allow 5000. Supply sufficient history.

Canonical symbol rules (bake in, no fetch needed)

Asset class	Form	Examples
Crypto	`{BASE}{QUOTE}` all-upper, no separator	`BTCUSDT`, `BTCUSDC`, `1000PEPEUSDT`
A-shares (`stock:cn`)	6-digit bare code (no `sh`/`sz` prefix, no `.SS` suffix)	`600519`, `300750`
Hong Kong (`stock:hk`)	5-digit with leading zero	`00700`, `09988`
US (`stock:us`)	Bare ticker	`AAPL`, `NVDA`
Forex (`forex:spot`)	6-letter `{BASE}{QUOTE}`	`EURUSD`, `USDJPY`

Always send canonical symbols, such as BTCUSDT. Do not send native symbols such as BTC-USDT-SWAP; use the discovery endpoints to confirm symbol availability.

Decision flow

When the user asks something data-related:

┌─ Is the user asking about LIVE data (current price, current indicator value)?
│   └─ YES → call api.mobiusquant.ai with Bearer token
│           - /api/klines for raw OHLCV
│           - /api/indicators for computed values
│           - /api/chart for klines + drawing protocol
│
├─ Is the user asking what's AVAILABLE (which indicators, which venues)?
│   └─ Have you fetched /llms-full.txt this session?
│       ├─ YES → answer from cache
│       └─ NO  → fetch /llms-full.txt once, then answer + cache
│
├─ Is the user asking HOW TO INTERPRET an indicator?
│   └─ Fetch /zh/indicators/<id>.md (Chinese — has signal interpretation + caveats)
│
├─ Is the user asking about a SPECIFIC SYMBOL's existence / canonical form?
│   └─ Call /api/symbols/search?q=<user's words> (engine resolves Chinese/English aliases)
│
└─ Is the user asking about PARAMETERS or OUTPUT COLUMNS of an indicator?
    └─ Fetch /indicators/<id>.md (English structured) OR /api/indicators/registry (live)

Caching guidance

/llms-full.txt, /indicators/<id>.md, /markets.md — content updates infrequently (when new indicators are added or venue list adjusts). Cache for the entire Skill session. Refresh on next session start.
/api/markets, /api/indicators/registry — live data, but content also rarely changes minute-to-minute. Safe to cache for 5-10 minutes.
/api/klines, /api/indicators (compute) — always live, never cache.

Quick example: user asks "show me MACD on BTC 1-hour"

Skill identifies intent: compute indicator on live data
Resolve symbol: user said "BTC" → from cached /llms-full.txt or by calling /api/symbols/search?q=BTC → canonical BTCUSDT on binance:perp
Resolve indicator: user said "MACD" → from cache: macd, default params {fast:12, slow:26, signal:9}

Call live:

bash

curl -X POST -H "Authorization: Bearer $MOBIUS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"exchange":"binance","market":"perp","symbol":"BTCUSDT",
       "interval":"1h","limit":100,
       "calc":[{"name":"macd","params":{"fast":12,"slow":26,"signal":9}}]}' \
  https://api.mobiusquant.ai/api/indicators

(Optional) If user asks "what does the histogram mean", fetch /zh/indicators/macd.md for interpretation.

Citation hygiene

When quoting from this docs site to the end user:

Always include the source URL so the user can dive deeper.
Prefer .md URLs over .html.
Mention if the source is Chinese-only (so user with English-only context knows).

This page is the single source of truth for Skill runtime data lookup. Skill maintainers: link to this page from your SKILL.md, do not inline-copy (linking lets you pick up updates automatically).

Skill Runtime Lookup Guide ​

TL;DR ​

Static reference (docs.mobiusquant.ai) ​

Bootstrapping: one-shot full reference ​

Per-indicator detail ​

Catalog & navigation ​

Live data (api.mobiusquant.ai) ​

Discovery endpoints (live, authoritative) ​

Natural-language symbol resolution ​

Compute / data endpoints ​

Canonical symbol rules (bake in, no fetch needed) ​

Decision flow ​

Caching guidance ​

Quick example: user asks "show me MACD on BTC 1-hour" ​

Citation hygiene ​

Skill Runtime Lookup Guide

TL;DR

Static reference (docs.mobiusquant.ai)

Bootstrapping: one-shot full reference

Per-indicator detail

Catalog & navigation

Live data (api.mobiusquant.ai)

Discovery endpoints (live, authoritative)

Natural-language symbol resolution

Compute / data endpoints

Canonical symbol rules (bake in, no fetch needed)

Decision flow

Caching guidance

Quick example: user asks "show me MACD on BTC 1-hour"

Citation hygiene