Finance & Investment Banking multi-agent-orchestrationmarket-researchportfolio-allocationcryptocurrencystock-analysisforexcommoditiesnextjs-dashboard

MAIA Investment Analysis (Tododeia)

Five AI agents research crypto, stocks, forex and commodities in parallel, adapt to your risk profile, and render a bilingual interactive market report — educational analysis, not financial advice.

FollowSkills review · FSRS-2.0
Use with care
55/ 100 5-point scale 2.8 / 5
1 2 3 4 5 6
1Trust16 / 25 · 3.2/5

Positives: strict read-only SKILL_DIR vs writable cache separation, mandatory educational disclaimer, XSS-hardened HTML fallback with JSON escaping, local-only runtime, standard optional env-var API keys. Deducted for: README promotes curl|bash one-line install (unreviewed supply-chain risk), unauthenticated dashboard, heavy reliance on runtime third-party web fetches with no integrity verification, and installer privileges not fully least-privilege.

2Reliability9 / 20 · 2.3/5

Positives: explicit stepwise workflow, single date capture, fallback ladders (primary→alternate→search→null), partial-failure rule, JSON retry, corrupted-history handling — error paths are unusually thorough. Deducted for: static review cannot reproduce key paths; quality of 5 sub-agent outputs depends on model behavior; no committed tests cover these paths, so end-to-end stability is unproven; capped below 10.

3Adaptability9 / 15 · 3.0/5

Positives: explicit trigger phrases, interactive risk-profile confirmation, declared non-fit boundaries (educational, not advice), HTML fallback without Node. Deducted for: core function depends entirely on overseas APIs (CoinGecko, Yahoo, Frankfurter) with doubtful mainland-China reachability and no mitigation disclosed; no Chinese-language support; broad triggers (e.g. 'investment advice') risk false invocation.

4Convention11 / 15 · 3.7/5

Positives: MIT license, version 2.1.0 with CI-enforced plugin./SKILL.md parity, referenced CHANGELOG, thorough bilingual README, JSON Schema data contract validated in CI. Deducted for: CI version-parity script appears to grep SKILL.md at repo root while the skill lives in a subdirectory (possible script defect), maintenance/update path rests on a single personal account, no FAQ or detailed known-issues list.

5Effectiveness6 / 15 · 2.0/5

Positives: claims a directly usable bilingual interactive report with sample fixtures, template, and schema; clear marginal value via automated multi-asset research. Deducted for: static review cannot verify outputs are directly usable; free-endpoint prices may lag; screenshots unverifiable here; cost/benefit of 5 parallel web-research agents unquantified.

6Verifiability4 / 10 · 2.0/5

Positives: a real CI workflow with lint, build, ajv schema validation, version parity and installer syntax checks — auditable primary material. Deducted for: no tests over the skill's key orchestration paths, no third-party execution evidence, historical accuracy self-described as a local heuristic rather than audited, thin coverage — capped below 5.

Evidence confidence:Low Reviewed Sep 10, 2026 Reviewed revision e57b7f8fec23
Before you use it
  • Core data sources (CoinGecko/Yahoo/Frankfurter) may be unreachable from mainland-China networks; no alternate endpoints or local caching are provided, so the whole chain may degrade to null values.
  • The README's curl | bash one-line installer executes an unreviewed script with user privileges; prefer manual clone and read install.sh before installing.
  • The dashboard has no authentication — never expose it publicly; output is educational analysis, not financial advice, and the historical-accuracy metric is a local heuristic, not an audited track record.
  • This was a static review with no execution: the 5-agent orchestration, fallback ladders and partial-failure rule are unverified in practice; run a small-scale trial first and verify data accuracy.
Review evidence [1][2][3][4]
See the full review method →

What does this skill do, and when should you use it?

This is the investment-analysis skill (v2.1.0) from the Hainrixz/maia-skill repository, a multi-agent investment research system for Claude Code branded "Tododeia." It asks your risk tolerance (conservative/moderate/aggressive), spawns four parallel sector research agents (crypto, stocks, currencies, commodities), then a strategy agent performs cross-sector synthesis, risk-adjusted ranking, and portfolio allocation, serving an interactive EN/ES Next.js dashboard locally. Market data comes primarily from free keyless APIs (CoinGecko, Yahoo Finance, Frankfurter) with optional Finnhub/Polygon premium keys; history is cached locally with accuracy tracking. An educational disclaimer is mandatory and output uses analytical phrasing rather than buy/sell instructions.

Loads five agent prompts from references/agent-prompts.md; asks the user's risk profile; fetches prices via keyless endpoints (CoinGecko / Yahoo v8 chart / Frankfurter) with a fallback ladder (alternate endpoint → WebSearch → null), using WebSearch for news and social sentiment; four sector agents return structured JSON under a numeric data contract; a strategy agent synthesizes macro environment, allocation, cross-sector insights, and risk-adjusted picks; report data is written to ~/.claude/cache/tododeia (keeping the last 30 runs) and a translation agent produces a Spanish report-es.; the primary path serves a Next.js dashboard at localhost:3420, falling back to a standalone HTML report served by Python's http.server on port 8420 when Node.js is unavailable.

  1. A retail investor who wants a daily or weekly panoramic market briefing across crypto, stocks, forex and commodities
  2. Spanish-speaking users who want both the UI and the report content to toggle between English and Spanish
  3. Traders who want scoring, position sizing, and allocation reshaped by a conservative, moderate, or aggressive risk profile
  4. Anyone who wants to track how previous AI signals performed (a local heuristic, not an audited record)
  5. Checking social sentiment and key news for an asset class before a trade
  6. Running in an environment without Node.js, using the standalone HTML report served via Python

What are this skill's strengths and limitations?

Pros
  • Parallel four-sector research plus a cross-sector strategy agent, with a numeric data contract validated in CI
  • Hybrid keyless data sources with a deterministic fallback ladder and an explicit partial-failure rule
  • Fully bilingual EN/ES — UI and report data translate on toggle while numbers and tickers stay intact
  • Strong security and privacy posture: CSP, XSS-hardened HTML fallback, local-only data cache, mandatory educational disclaimer
  • Accessible (ARIA, keyboard navigation, reduced-motion) and customizable agent prompts, risk multipliers, and styling
Limitations
  • Deeply dependent on Claude Code-specific mechanisms (Agent tool subagents, AskUserQuestion); porting requires rewriting the orchestration
  • Free public APIs may lag the market; some instruments fall back to search estimates
  • The "historical accuracy" metric is a local heuristic, not an audited track record
  • The dashboard has no authentication; the README explicitly warns against exposing it to the public internet
  • No evidence of a test suite or independent performance benchmarks in the source; requires internet — offline runs produce null prices

How do you install this skill?

Choose one of three ways: 1) One-liner: curl -sL https://raw.githubusercontent.com/Hainrixz/maia-skill/main/install.sh | bash; 2) Claude Code plugin: claude plugin install Hainrixz/maia-skill; 3) Manual: git clone https://github.com/Hainrixz/maia-skill.git, then ln -s "$(pwd)/maia-skill/.claude/skills/investment-analysis" ~/.claude/skills/investment-analysis and npm install --prefix maia-skill/dashboard. Optionally set FINNHUB_API_KEY or POLYGON_API_KEY for premium stock data. MIT licensed. Note: the repo bundles 2 skills; this profile covers only investment-analysis.

How do you use this skill?

In a Claude Code conversation, use a trigger phrase such as "Run an investment analysis", "Analyze the markets", "Run tododeia", or the Spanish "analiza mercados". Answer the risk-profile question; the five agents then run and the report opens at http://localhost:3420. On first open, acknowledge the educational not-advice notice and pick a language. Generated data goes to ~/.claude/cache/tododeia, not your project folder. Schedule recurring runs with /loop 24h /investment-analysis (daily) or /loop 168h /investment-analysis (weekly).

FAQ

Does it cost anything?
The skill is open source (MIT) and runs fully on free keyless APIs; paid Finnhub or Polygon keys are optional for better stock data only.
Can I use it without Node.js?
Yes. It falls back to a standalone HTML report built from assets/template.html, served by Python's built-in HTTP server on port 8420 (scanning 8420–8425). The fallback is single-language (no Spanish version).
What happens if data fetching fails?
A defined partial-failure rule applies: the failed sector's assets are excluded from picks, its allocation moves to cash with a warning; with no internet at all, a report with null prices and 'no data available' notes is still generated.
Can I treat this as investment advice?
No. The authors position it strictly as educational market research: a mandatory disclaimer precedes every report, language is analytical (Consider/Hold/Avoid), and users are told to consult a licensed advisor and assume all risk.

Related skills