cmc-skill — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited cmc-skill (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
A CoinMarketCap Agent Hub Strategy Skill (Track 2). Fuses 5 market signals into a 5-state regime classifier with an explicit trading posture per regime, backtested against buy-and-hold.
| field | value |
|---|---|
| name | market-regime-oracle |
| category | research / risk |
| type | Strategy Skill |
| tags | market, BTC, regime, risk, sentiment-data, market-data, derivatives-market-data, daily, strategy, backtest |
| track | BNB Chain AI Trading Agent Edition — Track 2 (Strategy Skills) |
| transport | MCP over stdio (Model Context Protocol) |
Reviews the daily crypto backdrop across momentum (RSI/MACD), sentiment (Fear & Greed), realized volatility, perp funding pressure, and exchange-flow pressure, then returns a conclusion-first market-regime read with a documented trading posture (target exposure + action) and a backtested context vs buy-and-hold.
Use it when you need a single, explainable answer to "what kind of market is this right now, and how much risk should I take?" — for daily risk-budgeting, sizing a BTC allocation, or layering a regime gate on top of other CMC skills (e.g. before acting on daily market overview or perp contract analysis).
Do not use it for live trade execution, order routing, wallet actions, or token launches. It is a research / backtest signal, not a trade instruction.
| name | type | default | notes |
|---|---|---|---|
symbol | string | bitcoin | CoinGecko coin id |
days | string | 365 | lookback window (CoinGecko free tier ≤ 365 days daily) |
force_refresh | bool | false | bypass the local data cache |
regime — one of RISK_ON, RANGE_BOUND, RISK_OFF, CAPITULATION, EUPHORIArecommended_posture — {target_exposure, action} for the latest daycomposite_score — fused signal score in [-1, +1]signal_scores — per-signal breakdown (5 signals)regime_distribution_12m — how often each regime occurredbacktest_vs_buy_hold — total return, max drawdown, Sharpe, Sortino, volposture_table — the full regime → exposure → action map| Regime | Target exposure | Action |
|---|---|---|
RISK_ON | 100% | Uptrend / accumulation — full exposure |
RANGE_BOUND | 40% | Sideways — light exposure, hold core |
RISK_OFF | 20% | Downtrend — defensive, raise cash to 80% |
CAPITULATION | 10% | Panic sell-off — max defensive, near-full cash |
EUPHORIA | 30% | Blow-off top — take profit, fade strength |
trigger or financial advice.
perp funding rates and exchange on-chain flows are not available from the authorized sources (CoinGecko, alternative.me), so those two signals are clearly-labeled proxies reconstructed from price/volume. They are not fake "real" data. See the repo README → Assumptions & proxies.
history without a demo API key, so the default backtest window is ~1 year.
are unreachable; it never fabricates market data.
The CMC MCP already exposes the raw ingredients this skill consumes (get_crypto_technical_analysis for RSI/MACD, get_global_metrics_latest for Fear & Greed, get_global_crypto_derivatives_metrics for funding, on-chain metrics for flows). This skill is the higher-order fusion + posture layer on top of those primitives.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.