Money Atlas Intelligence OS
Turns any financial market question into structured scenario analysis with explicit invalidation conditions — never a single price prediction.
SKILL.md itself declares requires_tools:false, explicit not_for boundaries (no live feeds, no order placement, no personalized advice), requires [UNVERIFIED] tagging, and mandates honest abstention over fabrication — a fairly complete least-privilege design with transparent data flow and no external side effects. Deductions: the shared repo contains execution_engine.py with Binance order placement and API key placeholders and data_feed.py calling external APIs; not scoring targets and unreferenced by the skill, but co-location creates misuse/confusion risk; publisher unverified; no rollback/confirmation mechanisms stated at the skill level.
SKILL.md is self-contained and tool-free, with an internal failure system (output invalid without alternative scenario/risk/uncertainty; abstention is a valid output). Deductions: static review only; the repo's tests target the Python reasoning engines (signal/risk/decision/fpc), while SKILL.md's output contract is backed only by structural validation; no execution evidence, so capped at 10 per anchors.
Clear trigger keywords, LIGHT/FULL modes, not_for boundaries and a handoff path for non-market questions. Deductions: trigger scope is extremely broad ('ANY financial question, ALWAYS trigger'), risking over-triggering; English-only with no Chinese-language support declared; mainland-China reachability unassessed (low impact since the skill needs no external services, but undeclared).
Apache-2.0 license, version 2.0, layered README/SKILL.md docs with progressive disclosure (references to full_mode.txt etc.), SECURITY.md, CI workflows and a validator script. Deductions: the referenced reference files and skill_config. were not present in the source material — hidden assumptions remain; no changelog; maintenance responsibility rests on a single individual.
Static cap 7. The framework is structurally complete (scenarios + confidence + invalidation), and a worked-examples directory is claimed but its content was not provided, so direct usability of outputs and marginal value over ordinary prompting cannot be verified. Deduction: core value claims lack auditable representative outputs.
Capped at 5. Real CI (compileall, pytest, skill validation, hygiene checks) plus targeted unit tests pinning key gates (no BUY on low confidence or distribution) are strong primary material. Deductions: tests cover the Python engines, not SKILL.md's declared behavior; no third-party corroboration; nothing was executed.
- This is a static source-only review; nothing was executed and confidence is low.
- The repo mixes the prompt-only SKILL.md with order-placement and market-data scripts containing API-key placeholders — do not conflate them, and never place real credentials in execution_engine.py.
- Files referenced by SKILL.md (full_mode.txt, light_mode.txt, architecture.md, system_prompt.txt, skill_config.) were not in the reviewed material; actual behavior may differ from claims.
- The skill produces analytical scenarios, not investment advice, and provides no live data; consult a licensed adviser for financial decisions.
- Trigger scope is very broad ('ANY financial question'); over-triggering is likely, and Chinese-language support is undeclared.
- Publisher is unverified (identity unknown); there is no changelog, so long-term maintenance is uncertain.
What does this skill do, and when should you use it?
This is a SKILL.md-based Claude skill for crypto, gold, forex, equities, and macro analysis. It ships two reasoning engines: the Genesis Protocol (First Principles + System Thinking) and a five-layer SMC market-structure model (accumulation through exit liquidity). The skill is pure prompt engineering — no tools or live data required — with optional stdlib-only Python modules in the repo. It explicitly refuses personalized investment advice and would rather abstain than invent a price when it has no data.
Runs the reasoning pipeline defined in SKILL.md: first deconstructs the question with the Genesis Protocol's five-question codex (what's happening, why, structural driver, asymmetric opportunity, per-scenario outcome), then maps micro → meso → macro → narrative system layers; next positions price within the five SMC layers, identifying the current layer, probable next move, and invalidation condition. Output is LIGHT mode (structure insight, key risk, takeaway) or FULL mode (situation map, bull/bear/base scenarios with entry/target/trigger, confidence and key unknowns). A constraint system flags any output lacking an alternative scenario or invalidation point, triggering a ⚠️ INSUFFICIENT EDGE re-evaluation.
- A trader asking 'should I buy, sell, or hold BTC right now' who wants a multi-scenario framework rather than a single call
- A macro analyst assessing how a Fed rate decision propagates through gold and dollar liquidity chains
- A novice investor who needs market-structure analysis with explicit invalidation points and risk warnings
- An independent analyst auditing consensus narratives to see who benefits and where institutional vs. retail positioning diverges
- A multi-asset investor wanting a portfolio-allocation and entry/exit decision framework while keeping the final call their own
What are this skill's strengths and limitations?
- Pure-prompt implementation: zero dependencies, zero cost, portable across models (frontmatter declares compatibility with any instruction-following model)
- Forces multi-scenario output with confidence levels and invalidation conditions, structurally resisting single-point predictions and overconfidence
- Built-in honest degradation: abstains and marks [UNVERIFIED] instead of fabricating numbers when data is missing
- Real test coverage: unit tests pin the critical constraints, validate_skill.py runs in CI
- Apache-2.0 with express patent grant; free for commercial use
- No live data integration — every concrete price level must be user-supplied or treated as unverified
- SMC (Smart Money Concepts) is a contested framework with little academic validation; output quality depends on its assumptions
- The Python modules are described as an optional execution layer; the README shows no quantitative backtest evidence independent of the prompts
- Documentation is English-first; non-English users must test trigger behavior themselves
- Not personalized investment advice and no order execution — it is a reasoning framework, not a signal service
How do you install this skill?
- Clone or download the repo: git clone https://github.com/ElmatadorZ/MoneyAtlas-ClaudeSkill-Agent
- Place the money-atlas-intelligence-os/ folder in your Claude skills directory
- Optional: run python tools/validate_skill.py to verify integrity and installability, and python -m pytest -q for the unit tests (after pip install pytest pyyaml)
No installation flow for other platforms is documented in the repo.
How do you use this skill?
The skill auto-activates on finance-related questions ('analyze gold right now', 'what will bitcoin do'); type FULL or LIGHT to override the output mode. Example prompt: 'FULL: analyze BTC bull/bear structure under a Fed rate-hike scenario.' Note the skill has no live price feed — supply current market data yourself if you want concrete levels, otherwise output will carry [UNVERIFIED] marks or explicitly state the gap.
How does this skill compare with similar options?
The repo situates it within its own Skynet skill ecosystem, explicitly routing non-market strategic questions to genesis-mind-strategic-intelligence; on buy/sell/hold questions it returns scenario frameworks and hands the decision back to the user — the fundamental difference from signal-calling services.