Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Quant & Fintech Code Auditor

    1

    Catch silent financial bugs in trading, DCF valuation & AI research codebases: 10^9 unit drift, percentage traps, look-ahead bias, and ungrounded LLM outputs.

    Free

    Security scanned
    Quant & Fintech Code Auditor

    Quant & Fintech Code Auditor

    Example session with this skill installed

    Audit routers/valuation.py and services/validators.py for unit mismatches, DCF calculation edge cases, percentage scaling traps, and Point-in-Time look-ahead bias.

    • Read your context and instructions
    • Compiled the quant & fintech

    Summary

    Audited routers/valuation.py and services/validators.py across all 5 quantitative guardrail layers. Found

    2 Critical mathematical/temporal defects and

    1 Warning that silently corrupt fair-value estimates and backtest integrity.

    Detailed Fixes

    1. Fix Billion-vs-Raw-Shares DCF Division (routers/valuation.py#L94)
    
    # Before: equity_value (Billion) / shares (raw) -> 0.000045
    # After: align dimensional scale explicitly
    scale_factor = 1.0 if shares_outstanding < 100_000 else 1_000_000_000.0
    fair_value_per_share = (equity_value * scale_factor) / shares_outstanding
    
    2. Replace Magnitude Guessing with Explicit Unit Metadata (routers/valuation.py#L118)
    def normalize_ratio(value: float, unit: str | None) -> float:
        if (unit or "").strip().upper() in {"%", "PERCENT", "PCT"}:
            return value / 100.0
        return value
    
    3. Enforce Strict Point-in-Time Boundary (services/validators.py#L244)

    Remove the unguarded fallback query (order_by(date.desc()).first() without date filter) and return INSUFFICIENT_DATA when no filing is published on or before as_of.

    Next steps

    • Apply the 3 patches above and re-run unit tests with boundary fixtures (0.8% growth, 125% growth, halted trading sessions, and pre-IPO dates).
    • Run quant-code-auditor on apps/web/ next to verify UI tabular-nums alignment and status enum contracts.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Generic code reviewers check syntax—they miss the financial bugs that silently ruin DCF valuations, leak future earnings into backtests, or let LLMs hallucinate numbers.

    Quant & Fintech Code Auditor applies an institutional 5-layer audit checklist distilled from real production equity research engineering (InvestIntel):

    1. Unit & Scale Consistency**: Catches billion-vs-raw-shares 10^9 division errors and dangerous if abs(x) > 1.0 percentage heuristics that inflate 0.8% growth into 80%.
    2. Point-in-Time (PIT) Integrity**: Hunts down look-ahead bias, unguarded fallback queries missing publication_date <= as_of, timezone mismatches (date.today() vs exchange time), and gapped-bar Beta distortions.
    3. AI Numerical Grounding**: Audits RAG/LLM pipelines to verify financial claims against retrieved database evidence and catch faulty scale multipliers.
    4. Backend & Cache Hygiene**: Detects FastAPI dynamic route shadowing (/{symbol} swallowing static routes), request-scoped DB sessions leaked to background tasks, and date-blind caching.
    5. Fintech UI Precision**: Checks for tabular-nums, frontend/backend contract enum mismatches, and NaN / zero-division guards in charts.

    Includes a complete anti-pattern reference library (references/fintech-bug-patterns.md) and a strict structured output format with line references and copy-paste fixes.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    Requires access to source code files. Does not perform live production monitoring or real-time trade execution. Assumes standard financial naming conventions for best results.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 5 days ago

    • Free with an Agensi account

    Requires

    Nothing to install.

    Permissions

    Read Files

    File Scopes

    **/*.py
    **/*.ts

    Read-only inspection of source files to audit valuation formulas, Point-in-Time queries, API routes, and UI numeric formatting. Never modifies files or runs shell commands.

    Listed5 days ago

    What's inside

    Frequently Asked Questions

    It targets "silent" financial errors like look-ahead bias in backtests, dimension mismatches (e.g., millions vs. billions), and temporal leakage that generic code reviewers often miss.

    The skill is optimized for Python (FastAPI/Quant stacks), TypeScript (React dashboards), and SQL, making it ideal for full-stack trading and fintech applications.

    No, this is a diagnostic tool for development environments; it reviews source code and research pipelines but does not execute live trades or monitor production traffic.

    You receive the full skill package including the 5-layer audit logic, custom fintech bug signatures, and structured reporting templates for your AI agent.

    The skill identifies "future data" leakage by verifying your date-filtering logic, timezone alignments, and session-aware caching to ensure backtest results are statistically valid.

    Simply provide the agent with access to your source files or code snippets; it follows a 4-step workflow to map data flow and generate a prioritized audit report.

    New in Code Quality & Review

    Free