Monetary Sentinel
Reads macro and on-chain liquidity data to classify the market as Risk-On or Risk-Off, returning a liquidity score with cited reasoning.
Read-only macro-analysis agent with no external writes or credential needs; data sources (FRED, DefiLlama, Fear & Greed) are disclosed and a no-look-ahead ('never later prints') constraint is explicit, giving high data-flow transparency. Deducted for: no source-verification/hash disclosure for external feeds, and insufficient user-facing notice of degraded behavior when silently falling back to OHLCV proxies.
SKILL.md and persona.md output contracts are consistent (tier0/v1, macro_regime_state, Liquidity_Score); few-shot examples are self-consistent and fallback paths for missing files/endpoints are declared; the repo has CI and a pytest suite. Deducted for: static review cannot reproduce key paths, and failure-feedback quality on abnormal input is unverified — capped at 10 per anchors.
Trigger conditions are clear (macro regime / liquidity crisis queries), the scenario (BTC crypto macro regime classification) and output contract are well defined. Deducted for: no explicit non-fit boundaries or capability limits; core data sources (FRED, DefiLlama, LLM APIs) are overseas services with questionable mainland-China reachability — pinned CSV fallback mitigates but live-mode dependence remains.
SKILL.md + persona.md layering is clean; README thoroughly documents the data layers, MANIFEST hashes and the AGPL-3.0 license; repo-level maintenance signals (CI, SECURITY.md) exist. Deducted for: no skill-level version number or changelog, no FAQ or known-limitation section, maintenance ownership unclear (publisher unverified).
Goal is clear — output a macro regime state plus liquidity score with checkable reasoning, which beats manually aggregating several feeds. Deducted for: static review cannot verify output correctness or regime-classification quality; the reported backtest returns attach to the whole strategy, not this skill, so comparative-benefit evidence is limited — kept below the static cap.
There is a real CI workflow, a pytest suite (with nexus-context assertions), pinned data/ with MANIFEST hashes, and good fact/inference separation. Deducted for: tests cover context construction rather than this skill's core judgment path, and there is no third-party execution evidence — capped at 5 per anchors.
- This is a static source-only review; nothing was executed and confidence is low.
- Live mode depends on overseas services (FRED, DefiLlama) and an LLM API; mainland-China reachability is unverified — offline use requires the repo's pinned CSVs.
- Silent fallback to an OHLCV liquidity proxy when sources are missing degrades regime quality; documentation does not adequately flag this.
- Macro thresholds (e.g., VIX≥32, TVL 7d≤−8%) are hard-coded heuristics without methodological validation or backtest attribution; do not treat outputs as trading advice.
- Publisher identity is unverified and skill-level versioning/maintenance path is unclear; verify the latest commits before use.
What does this skill do, and when should you use it?
Monetary Sentinel is a skill inside olaxbt/ai-market-maker (AGPL-3.0), located at src/agents/operator/1.1_monetary_sentinel/SKILL.md. It synthesizes the Fear & Greed index, FRED macro indicators (VIX, fed funds, 10y yield, trade-weighted USD, SPX, WTI), and DefiLlama stablecoin supply plus all-chain TVL to classify the macro regime as Risk-On or Risk-Off and produce a liquidity score. Data is strictly read as-of the bar date with no later prints, making it backtest-safe. It does not execute trades; it is a macro-state judgment layer meant to feed larger trading systems.
The skill exposes analyze(ticker, market_data, nexus_context) as a Tier-0 contract and answers direct queries like "What's the current macro regime for BTC?" or "Is there a liquidity crisis signal?". It reads Fear & Greed (0–100), FRED data (VIX, fed funds, 10y yield, trade-weighted USD, SPX, WTI), and DefiLlama lag-1 stablecoin supply and all-chain TVL with 7d change; when missing, it falls back to OHLCV window return/vol as a liquidity proxy. Interpretation rules: VIX ≥32 or F&G ≤25 → Risk-Off; rising fed funds or strong DXY → tighter dollar liquidity, lean Risk-Off; stablecoin or TVL 7d ≤ −8% → on-chain drain, Risk-Off; F&G ≥75 with falling VIX and expanding stablecoins → Risk-On. Query via /monetary_sentinel?ticker=BTC/USDT, which returns the macro regime state, liquidity score, and reasoning.
- A quant trader checks whether BTC is in a Risk-Off macro regime before sizing a position
- A backtest researcher needs a macro overlay aligned to historical bar dates to avoid lookahead bias
- A crypto fund team monitors stablecoin supply and TVL 7d changes to detect on-chain liquidity drains
- A developer building on the ai-market-maker multi-agent pipeline needs a macro-state upstream node
- A macro analyst asks a direct question like "Is there a liquidity crisis signal?" and wants reasoning citing concrete printed indicators
What are this skill's strengths and limitations?
- Explicit data sources (Fear & Greed, FRED, DefiLlama) with a strict as-of-bar-date, no-later-prints discipline that is backtest-safe
- Transparent, auditable interpretation rules (e.g., VIX ≥32 or F&G ≤25 → Risk-Off) with mandatory citation of printed values
- Multiple fallbacks: OHLCV proxies when macro CSVs or live endpoints are missing, reducing environment dependence
- Clean dual interface: structured Tier-0 analyze contract plus natural-language direct queries
- It does not execute trades — output is only regime state and score, so it must be paired with other components
- Depends on external data sources (FRED, DefiLlama, Fear & Greed) for availability and timeliness, and DefiLlama data is lag-1
- Fixed thresholds (−8%, 32, 75) are simple rules; the source provides no evidence they were validated statistically
- No standalone installation or test procedure is documented for this skill in isolation; setup depends on the whole monorepo
How do you install this skill?
The skill ships as part of the ai-market-maker monorepo: git clone https://github.com/olaxbt/ai-market-maker.git, install uv, run uv sync --extra dev, and configure .env (Binance Testnet and OpenAI keys optional per README). Optionally run the full platform with docker compose -f docker-compose.prod.yml up --build -d. The skill file lives at src/agents/operator/1.1_monetary_sentinel/SKILL.md. Steps to install this sub-skill standalone as an OpenClaw skill are not documented in the source.
How do you use this skill?
In backtest mode it relies on data/macro/*.csv via HistoricalNexusProvider; otherwise it uses the live Nexus market_overview. If those files/endpoints are missing, it falls back to OHLCV stubs. Example query: /monetary_sentinel?ticker=BTC/USDT returns macro regime state + liquidity score + reasoning. You can also ask directly: "Is there a liquidity crisis signal?". When reasoning, quote the printed VIX / DXY / fed funds / 7d TVL change and never invent later data.
How does this skill compare with similar options?
The source names no direct competitor. Within the same repo it complements the research desks (Market Scan, Technical TA, Statistical Alpha, Sentiment), owning the macro-regime dimension while they cover theirs.