// the find
himself65/trade-skills
A collection of skills for AI trading.
A Claude Code plugin that bundles a single 'skill' for options trading: a markdown knowledge base of 37 named pitfalls plus ticker-specific case studies (INTC, SNOW, MU, etc.) that an agent reads before answering trading questions. It's aimed at developers who already use Claude Code/agent skills and want an LLM to reason about options structures with some guardrails against common mistakes.
The eval suite under plugins/trade/evals is the real differentiator — case.yaml scenarios with graders checking specific behaviors (does it cite the right pitfall, does it avoid a jade lizard recommendation, does it route to the daily-check skill) instead of just shipping a prompt and hoping. The ticker/ writedowns ground the abstract pitfalls in dated, specific trades rather than generic platitudes, so you can tell this is actually used, not written speculatively. Frontmatter + index.md graph structure (OKF) keeps the knowledge base navigable and diffable in version control instead of being one giant prompt blob.
Labeled JavaScript but there's barely any code — this is markdown content shipped through npm tooling, so 'stars/language' signals are misleading about what you're actually getting. There's no way to verify the trading content itself is sound; the evals test plumbing (did the right file get read, did the right skill fire) not whether the financial reasoning is correct, so a confidently-wrong pitfall would pass every grader. No backtesting or outcome tracking tying any pitfall back to actual P&L — it's a personal opinion set wrapped in agent-skill infrastructure, and that infrastructure doesn't make the opinions more correct. OKF itself is a one-person convention borrowing a Google-published name, not an adopted standard, so 'portable, vendor-neutral' is aspirational more than true today.