How to use the Commodity Analysis agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a request matches the skill's description. It changes how the agent reasons about a topic rather than adding live data.

The commodity-analysis skill covers six areas: crude oil supply-demand balance, gold pricing, copper as an economic predictor, inventory cycles, futures premium and discount structure, and seasonality. It focuses on crude oil, gold and copper.

• Supply-demand: OPEC production, US shale output, rig count, Strategic Petroleum Reserve, plus demand indicators such as China crude imports and global PMI.
• Gold: a four-factor model covering real rates, the US dollar index, safe-haven demand and central-bank buying.
• Copper: leading-indicator logic and inventory thresholds for LME and SHFE.
• Futures structure: contango versus backwardation and roll yield.
• Output: a report with directional signals written so they can be tested against historical data.

Install the commodity-analysis skill

Download the skill ZIP from this page and unzip it. The skill is a folder that contains SKILL.md. Place that folder in one of these locations:

• Personal skills folder: ~/.claude/skills/
• Project-specific: the .claude/skills/ folder inside your project
• Claude apps: upload the ZIP under Settings → Capabilities → Skills

Trigger the skill with natural requests

You do not need a special command. Describe the commodity and the question in plain language, and the agent will apply the framework when the request matches the skill's description. Naming the commodity (oil, gold or copper) and the kind of view you want helps.

Supply the market data yourself in the request or in a file. The skill describes where to find data, such as OPEC and EIA reports, LME and SHFE inventories, and customs data, but it does not fetch prices.

• "Using the commodity-analysis approach, assess the current oil supply-demand balance. Here are the latest EIA weekly figures: ..."
• "Apply the gold four-factor model to these real rate and dollar index readings and give a directional view."
• "Check the copper futures curve shape and tell me whether backwardation confirms the demand picture."
• "Give me a commodity analysis report for copper with a composite score and bias."

Example workflow: a crude oil report

The skill's five-step method runs in order: supply-demand sets direction, inventory sets rhythm, term structure confirms, seasonality overlays, and macro checks validate the view. The output follows a standard report layout with sections for supply-demand, inventory cycle, term structure, a composite score table and a trading direction.

The composite score uses a -2 to +2 scale per dimension, so a total can be read as a simple bias. The skill's own example gives +3 out of 5 (+0.6) as a bullish bias that is not a strong signal.

What you need before you start

Python is needed only if the agent should run the composite scoring or term-structure code. Without it, the agent can still reason through the framework in text.

You also need market data from your own source or another tool. The skill names OPEC, EIA, LME, SHFE, customs and similar sources. It does not specify API keys or accounts, so check what each source requires.

• Python (optional, for running the scoring and term-structure code)
• Market data you supply, or from another tool
• No API keys or accounts are specified by the skill

Safety notes: scripts and network access

The skill is listed as not running scripts and not using the network. The code blocks in SKILL.md are illustrations the agent may adapt. It does not fetch live prices or inventory numbers.

The thresholds and rules are simplified illustrations, not validated trading strategies. Treat any signal as a hypothesis to test, not an instruction to trade.

• No scripts are run by the skill as listed.
• No network use is listed, so any live data must come from you or another tool.
• Do not trade real money on these signals without independent verification.

Tips for better results

Give the agent recent, specific numbers rather than asking it to recall current data. The framework depends on the inputs, and it can only check thresholds you provide.

Ask for the reasoning as well as the signal, so you can see which of the four dimensions drove the view.

• Provide dated data with its source, such as the EIA weekly report date.
• Ask for each dimension's score and basis, not only the final bias.
• Use the output to build a backtest, then test the rules on historical data yourself.