clawock — Bull/Bear Investment Decision Desk

    1

    **Research & debate** — the agent gathers evidence with its own tools, then must write both sides: a bull case and a genuine bear case, each citing specific evidence rows.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    clawock — Bull/Bear Investment Decision Desk

    Example session with this skill installed

    Run an investment decision for NVDA US. I have the Q3 filings in the workspace. Research the current valuation and provide a bull/bear debate before deciding on a watch action.

    • Read your context and instructions
    • Compiled the clawock

    I have completed the research and debate for NVDA (US/USD). Supporting evidence from the Q3 filing was reconciled against market valuation risks. The decision has been published to the workspace.

    Decision Summary

    Subject: NVDA (US/USD) | Decision ID: nvda-2026-08-08

    ComponentContent
    Bull CaseData center revenue growth exceeds 90% YoY; strong primary filing support.
    Bear CaseForward P/E sits at 2.8 standard deviations above 5-year mean; market class risk.
    ThesisFundamental strength is offset by extreme valuation multiples.
    Actionwatch
    InvalidationForward P/E reverts to mean or data center growth dips below 50%.
    {
      "schema_version": 1,
      "workflow": {"id": "investment-decision", "version": "1.1.0"},
      "decision_id": "nvda-2026-08-08",
      "as_of": "2026-08-08T10:00:00+00:00",
      "subject": {"ticker": "NVDA", "market": "US", "currency": "USD"},
      "evidence": [
        {"id": "q3-filing", "stance": "supporting", "summary": "Revenue grew 94% YoY.", "source": "Q3 Filing", "source_class": "primary", "observed_at": "2026-08-08T09:00:00+00:00"},
        {"id": "valuation-peak", "stance": "opposing", "summary": "P/E ratio at historical highs.", "source": "Market Snapshot", "source_class": "market", "observed_at": "2026-08-08T09:30:00+00:00"}
      ],
      "debate": {
        "bull_case": {"summary": "Strong revenue growth.", "evidence_ids": ["q3-filing"]},
        "bear_case": {"summary": "Valuation risk is high.", "evidence_ids": ["valuation-peak"]}
      },
      "thesis": {
        "statement": "Growth is robust but valuation is stretched.",
        "confidence": 0.65,
        "invalidation_conditions": ["Growth slows below 50%"]
      },
      "decision": {
        "action": "watch",
        "rationale": "High valuation blocks entry despite growth.",
        "evidence_ids": ["q3-filing", "valuation-peak"],
        "order": null
      }
    }
    

    Next steps

    • Review the receipt in .clawock/runs/ to confirm the certificate pin.
    • Set a price alert for the identified invalidation condition.
    • Run clawock record if you decide to override this with a manual entry.

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

    About this skill

    What it does

    clawock turns "should I buy this stock?" into an auditable decision instead of a vibe. When you ask your agent for an investment decision, this skill runs a fixed contract:

    1. Prepare — the clawock engine assembles a certified request: the decision contract plus context documents with sha256 fingerprints and evidence gates (minimum supporting evidence, minimum opposing evidence, a confidence cap when no primary source is cited).
    2. Research & debate — the agent gathers evidence with its own tools, then must write both sides: a bull case and a genuine bear case, each citing specific evidence rows. A strawman bear case fails validation.
    3. Decide — a thesis with a confidence level and explicit invalidation conditions, plus a bounded action (buy / watch / reject / …). Any order must reconcile to the cent: quantity × unit price × FX.
    4. Publish & settle — Python validates the artifact against the contract (evidence linkage, debate linkage, money identities, timestamps) and issues a receipt. The model can never grade itself; outcomes are scored later against market prices by code, not by the agent.

    Who it is for

    Investors and builders who use AI agents (Claude Code, Cursor, Codex, DeepSeek Harness and other SKILL.md-compatible harnesses) for stock research on HK and US markets, and who want the reasoning — including the case against — kept on the record and scored afterwards.

    Requirements

    • Python ≥ 3.11 and the free, open-source clawock engine: python -m pip install clawock (MIT licensed, github.com/KCNyu/clawock). This listing is the official skill published by the clawock author; the engine stays free and open source.
    • A harness that can read files, run shell commands and do web research.

    What is included

    • SKILL.md — the full decision workflow (prepare → research/debate → publish, self-check list, rejection-repair table, conversation-verdict recording).
    • references/decision-contract.md — the readable decision contract.
    • references/decision.schema.json — the JSON Schema for decision artifacts.
    • assets/ — worked examples: a decision, an order-carrying decision whose money identities reconcile, and an outcome record.

    What is NOT included

    • Not financial advice and not an auto-trader: clawock never places orders; it produces validated decisions and receipts. You place your own orders.
    • The DeepSeek Harness GUI plugin (decision-trace tab) is distributed separately via npm as clawock-dsh; it is not part of this skill package.
    • No market-data subscription: the agent uses its host's own research tools.

    Live proof, updated daily from a real-money HK + US desk run by the author: https://kcnyu.github.io/clawock/

    Requires Python 3.11+. The model cannot self-grade; all decisions must be settled by the external clawock Python CLI. Validates logic and math, not actual market profitability.

    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 any payment.

    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 forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 2 days ago

    • Free with an Agensi account

    Requires

    Python ≥ 3.11
    Free open-source clawock engine (pip install clawock)
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    Standard prompting often suffers from confirmation bias and math errors; this skill enforces the Clawock Protocol, which mandates a formal bull/bear debate and uses external Python validation to ensure every trade is backed by verified data.

    This skill requires Python 3.11+ and the Clawock CLI, and it is specifically optimized for the DeepSeek Harness environment to ensure rigorous data integrity and schema compliance.

    Each purchase includes the 'Investment Decision' skill logic, the required JSON schemas for decision artifacts, and integration templates for the market, primary, and secondary source classes.

    Yes, the skill is designed to log every decision to a shared ledger, specifically tracking emotional triggers like FOMO or revenge trading to provide a verifiable audit trail for compliance or review.

    The 'clawock' library and CLI act as an external judge, running Python-based arithmetic checks on order quantities and FX conversions that the LLM cannot bypass or self-grade.

    The skill utilizes an 'as_of' timestamp mechanism, ensuring the agent only considers evidence available up to that specific moment to maintain the integrity of the research window.

    New in Finance & Accounting