About this skill

Problem

Portfolio-focused agents often let LLMs estimate holdings, drawdown, duration, and rebalance figures, which makes results hard to audit. investorclaw separates deterministic analytics from narrative synthesis: it first computes weights, performance, Sharpe ratios, and sector breakdowns from broker holdings files, then generates natural-language answers with cited numbers. It is useful for portfolio review with stable, traceable math, not general finance Q&A.

How it works

portfolio_ask is the primary entry point, auto-loading holdings, refreshing market data, and returning a narrative with referenced figures.
Tools such as portfolio_initialize_status, portfolio_setup, and portfolio_refresh handle bootstrapping, discovery, and forced refreshes.
The ic-engine backend runs in a standalone container and exposes MCP-HTTP and REST endpoints for FRED yield curves, bond duration, scenario rebalancing, and response history.
Responses can include an HMAC signature to confirm that numbers came from the deterministic engine rather than model fabrication.

Boundaries

It requires broker-exported CSV, Excel, or PDF holdings files; without one, it returns a missing-file notice.
TOGETHER_API_KEY controls narrative quality; without it, deterministic calculations still run but prose degrades.
Large portfolios on free quote sources may be rate-limited; richer results may need keys such as Finnhub, FRED, or Massive.
It is portfolio-specific and should not be used for generic finance concepts without holdings context.

Screenshots

Use Cases

Import a broker export, load the portfolio, and ask for returns, Sharpe ratio, and sector weights to produce an auditable summary.
Before a monthly review, query cash weight, bond duration, and FRED yield-curve context to check whether the current mix follows target allocations.
Test a rebalance scenario for a 200-symbol portfolio and compare expected weight changes under different target allocation assumptions.
Review a stored portfolio answer with response history tools, flag a bad run, and keep the record for later audit or debugging.

Best For

Personal portfolio managers who want agents to answer returns, Sharpe ratios, and sector allocations from broker holdings files.
Research analysts doing portfolio diligence who need bond duration, yield-curve context, and rebalancing scenarios instead of model guesses.
Financial-agent engineers who want to call portfolio analytics over MCP/REST and integrate signed results into workflows.
Compliance reviewers auditing AI finance output who need response history, bad-response flags, and evidence that numbers came from deterministic math.

Related Skills

Professional Contract Review

@user_c3030ed3/accurlex-contract-review

Reviews Chinese-law contract text via the accurLex API, returns the full review opinion, and exports a Word document.

Professional 2026.08.29 ↓ 394

Active Comparator Single-SOC FAERS Safety Comparison Planner

@user_c7479c7a/active-comparator-single-soc-faers-safety-comparison

Generates executable FAERS active-comparator safety comparison research plans focused on one SOC or PT family.

Professional 2026.08.29 ↓ 398

Financial Data Analysis Expert

@user_15292d5a/yjkj-finance-data-analysis

A reference framework for financial data analysis covering statements, KPIs, visualization, and reporting templates, without executing code or collecting data.

Professional 2026.08.29 ↓ 362

Device Selection Advisor

@user_7a5540f4/device-selection-advisor

A knowledge-base-driven advisor for weak-current, network, and telecom equipment selection, covering parameter comparison, model recommendations, compatibility checks, and budgeting.