Agent Workflows

Use Option Desk with an agent

Guide home | Installation | Examples

For manual, shell, npx, or browser ZIP installation, start with Install skills in your agent.

Install the local tools before the local plugin. For a browser agent, use the separate hosted connection.

Claude Code

Run these commands inside Claude Code:

/plugin marketplace add Iman/agent-driven-options-desk-and-skills
/plugin install option-desk@option-desk

The local plugin includes skills, commands, two reviewer agents, and an MCP declaration. The reviewer agents are desk-data-auditor, which checks the freshness and consistency of the artifacts on disk before anyone reports from them, and options-risk-reviewer, which re-derives the risk of a chosen structure and states what would have to be true for it to lose. The declaration starts optiondesk-mcp, which must be on the runtime's PATH. See the official plugin instructions.

Codex

Run these commands in a terminal:

codex plugin marketplace add Iman/agent-driven-options-desk-and-skills
codex plugin add option-desk@option-desk

Codex can also discover the repository's .agents/skills without a plugin. The full local installer supplies the calculation tools. The OpenAI plugin guide describes the package format and marketplace workflow.

First local prompt

Open an agent session in the cloned repository, then ask:

Use examples/chain-synth.json, the supplied synthetic teaching sample.
I can use this file for private analysis.
Save this walkthrough under artifacts/tutorial.
Validate the chain, calculate its Greeks and positioning, and build an iron condor.
State the source, valuation date, missing inputs, and degradation before the numbers.
Show the payoff and explain the limits of the model.

Expected: the agent uses the supplied file and reports synthetic input. A response that calls this live market data is incorrect.

Ask by task

Task	Example prompt	Local skill
Setup	Install the tools and check why MCP is unavailable.	desk-setup
Contract risk	Show delta, gamma, theta, and vega from this saved chain.	options-greeks
Chain positioning	Show gamma walls and state the dealer-sign assumption.	options-positioning
Structure comparison	Compare an iron condor with a straddle from this chain.	options-strategy
Forward distribution	Simulate the underlying from permitted price history and report convergence.	options-simulation
Historical research	Backtest the structure and explain the benchmark and omitted costs.	options-backtest

Repeat a local workflow

The Claude Code plugin supplies these commands:

Command	Purpose
/desk-open SPY	Retrieve a permitted chain, calculate positioning, and compare structures.
/desk-risk SPY 30	Simulate the underlying and review the risk assumptions.
/desk-test SPY iron_condor	Research a structure with backtest and forward-test workflows.
/desk-mark	Mark open paper positions against saved chains.
/desk-watch SPY	Report material changes since the previous run.
/desk-complete SPY	Work toward a defined set of complete artifacts.

Omitted second arguments take defaults: /desk-risk uses fourteen business days, /desk-test uses five years, /desk-open uses the nearest listed expiry, and /desk-complete uses the nearest listed expiry beyond a week. Provider commands require an enabled data source. Use Loops for budgets, stop conditions, and scheduling limits.

Browser agents and the hosted service

Follow the separate ChatGPT steps or Claude chat steps. They include individual skill uploads, account requirements, and the MCP connection.

The hosted MCP endpoint is:

https://optiondesk.avidquant.com/mcp

Add it through your runtime's supported custom MCP or plugin setup. Connection availability depends on your account and workspace configuration. A directory submission does not establish that every user can install the listing.

Start without an upload:

Show the SYNTH market chart and Greek ladder.
Identify the data as synthetic and explain what each chart measures.

Then try a structure:

Show the payoff for a SYNTH iron condor.
Explain the legs, breakevens, maximum gain, and maximum loss.

For your own chain, attach only a permitted CSV or JSON file. Name its source and state that you can send it for private analysis. Ask for validation before calculations.

The hosted workflow supports snapshot analysis, Greeks, positioning, and strategy plots. It does not fetch live chains or offer the local simulation and backtest workflow. The hosted privacy policy governs uploads.

Use one local or hosted Option Desk connection per session to avoid duplicate tool names.

Next: Import your data or read the results.