Pro Bias Analyst
Determines whether smart money is buying or selling an asset like BTC by fusing ETF flows, futures basis, and options skew into a single pro bias score.
Read-only analytical skill with no external side effects; data sources (Nexus ETF endpoints, futures basis, options skew) disclosed, and basis-only fallback when Nexus is disabled is documented; but no least-privilege statement, confirmation mechanism, or rollback notes, and NexusDataClient implementation is not visible in the evidence provided.
SKILL.md and persona.md output contracts (tier0/v1, Pro_Bias/ETF_Trend/ema_slope) are mutually consistent; the repo ships pytest suites, deterministic fixtures (including etf_metrics), and CI; however no skill-specific test evidence for the pro_bias computation or basis-only fallback exists, and static review cannot verify reproduction, so capped at 10 and deducted.
Scenario (crypto institutional flow / ETF trend assessment) and trigger phrasing are clear, with documented degradation when Nexus is disabled; but capability boundaries and non-fit ranges are undeclared, and core function depends on Nexus/overseas APIs with no mainland-China reachability discussion, so deducted.
Reasonable doc layering (SKILL.md + persona.md), complete output contract and few-shot examples, plus repo-level AGPL-3.0 license, CI and SECURITY.md; but the skill files themselves lack versioning/changelog, known-limitation disclosure, and an explicit maintenance owner; publisher identity unverified, so deducted.
The described capability has a corresponding desk role and test fixtures in the repo architecture, but no verifiable representative outputs are provided and the marginal benefit claim is documentation-only; static review cannot exceed 7, hence the score.
Repo tests, CI workflow and deterministic fixtures corroborate the agent system's existence, but direct test evidence for this skill's path is missing, few-shot results are author claims not independently reproducible, and fact/inference separation is thin, capped at 5 and deducted.
- Static review only; no code was executed. All conclusions are based on source files.
- Core function depends on Nexus data endpoints; mainland-China reachability is undocumented — verify before use.
- The NexusDataClient implementation and pro_bias scoring logic are not visible in the evidence; actual numeric behavior cannot be verified.
- Output feeds trading signals; research reference only, not investment advice.
- Publisher identity is unverified; maintenance and update path unclear.
- Skill files carry no version number or changelog; upgrades may introduce breaking changes.
What does this skill do, and when should you use it?
Pro Bias Analyst is one of 14 skills bundled in the olaxbt/ai-market-maker (AIMM) repository, located at src/agents/operator/3.2_pro_bias_analyst/SKILL.md. It answers one specific question: which direction is institutional/professional money leaning. The skill reads three data sources — Nexus ETF endpoints, perpetual-vs-spot futures basis premium, and 25-delta risk reversal options skew — and returns a pro bias score, ETF trend, and EMA slope via a query interface like /pro_bias_analyst?ticker=BTC/USDT. When Nexus is disabled, it gracefully degrades to basis-only analysis. It fits crypto traders who already run, or are willing to deploy, the full AIMM stack.
The skill exposes an analyze(ticker, market_data, nexus_context) call following a Tier-0 contract, and supports direct natural-language queries such as 'Is smart money buying or selling BTC?' and 'What's the ETF flow trend?'. It calls Nexus ETF endpoints for flow data, reads the futures basis (perpetual vs spot premium), and fetches options skew measured by the 25-delta risk reversal, then returns a pro bias score, ETF trend, and EMA slope. It does not execute trades — it produces a directional read only. It depends on NexusDataClient and falls back to basis-only analysis when Nexus is disabled.
- A crypto trader wants institutional-direction confirmation before opening a position, asking 'Is smart money buying or selling BTC?' to calibrate long/short bias.
- An ETF-focused researcher wants a quick read on BTC/ETH spot ETF flow trends without stitching together multiple data sources manually.
- A quant team running the AIMM multi-agent system uses this skill as the pro-bias input node in its Tier-0 consensus workflow.
- In environments where Nexus data is unavailable, a trader still gets a rough directional read from the basis-only fallback mode.
- An options trader wants 25-delta risk reversal skew as a corroborating sentiment signal.
What are this skill's strengths and limitations?
- Tightly scoped to one clear task: determining professional money direction, outputting a structured score rather than vague commentary.
- Multi-signal fusion: ETF flows, futures basis, and options skew provide independent cross-validation.
- Graceful degradation: basis-only fallback keeps the skill usable when Nexus is unavailable.
- Follows the repo's standardized Input → Process → Output → Feedback Tier-0 contract, making it easy to slot into the AIMM multi-agent pipeline.
- Depends on the entire AIMM stack (Python 3.11+, uv, TA-Lib, Docker, etc.) — no lightweight standalone install exists.
- Full functionality requires Nexus Skills API access; its cost and availability are not detailed in the docs.
- The README's example backtest numbers reflect default configuration of the whole system, not independent validation of this skill; a single profitable backtest is explicitly not proof of edge.
- No skill-specific test evidence or accuracy metrics are provided, so score quality cannot be independently assessed from the source material.
How do you install this skill?
The skill cannot be installed standalone; install the whole olaxbt/ai-market-maker repository:
- git clone https://github.com/olaxbt/ai-market-maker.git && cd ai-market-maker
- pip install uv && uv sync --extra dev
- Install TA-Lib (recommended: conda install -y ta-lib -c conda-forge)
- cp .env.example .env and fill in keys (optionally Nexus Skills API access)
- The skill file itself lives at src/agents/operator/3.2_pro_bias_analyst/SKILL.md and is available as part of the repo. No standalone install steps for just this skill are documented.
How do you use this skill?
After installing the repository, query via the interface: /pro_bias_analyst?ticker=BTC/USDT, which returns pro bias score + ETF trend + ema slope. You can also invoke it as analyze(ticker, market_data, nexus_context) within the agent workflow, or ask directly in natural language, e.g. 'Is smart money buying or selling BTC?' or 'What's the ETF flow trend?'. With Nexus enabled it uses full data; with NEXUS_DISABLE=1 it automatically degrades to basis-only analysis.