How to use the Investor Panel agent skill (investor-panel)

What the investor-panel skill makes the agent do

An agent skill is a folder of instructions and reference files that an AI agent loads when your request matches the skill's description. This one is a multi-persona voting panel. Its description covers nine investing schools, including classic value, growth, macro hedge, technical trend, Chinese value, A-share speculative trading, quant systems, tech leaders and AI positioning.

For each persona, the agent is asked to return a structured result with the fields signal, confidence, score, verdict and comment, plus reasoning, pass and fail lists. The skill then tallies the results into a consensus figure and vote and signal distributions.

The skill works from pre-collected data. It reads a dimensions.json file and a raw_data.json file for one ticker and does not fetch live market data itself.

Install the skill from the ZIP file

Download the ZIP for investor-panel from the directory page and unzip it. Then place the resulting investor-panel folder in one of two locations: your personal skills folder for all projects, or a project-specific folder for one codebase.

Alternatively, in the Claude apps you can upload the skill under Settings, then Capabilities, then Skills.

• Personal install: put the folder in ~/.claude/skills/
• Project install: put the folder in the project's .claude/skills/ directory
• App install: Settings → Capabilities → Skills, then upload the skill

Prepare the input files

The skill expects you to supply the stock data. For a ticker, the files go in a cache folder named after the ticker. The source describes reading the 19-dimension scores and the raw data, and writing its output to a panel.json file in the same folder.

Several supporting files are referenced by the skill but are not included in the file list for this package. Those are the Python modules scripts/lib/investor_db.py and scripts/lib/seat_db.py, and references/quotes-knowledge-base.md. If they are missing, the skill may fail or the agent may improvise quotes, so check for them before you rely on the output.

• Required: .cache/{ticker}/dimensions.json
• Required: .cache/{ticker}/raw_data.json
• Referenced but not in the listed files: scripts/lib/investor_db.py, scripts/lib/seat_db.py, references/quotes-knowledge-base.md

Trigger the skill with natural requests

The skill's description lists trigger phrases in Chinese. Matching the wording helps the agent pick the skill. The phrases are roughly: a panel review, what the 65 experts think, would someone buy a given stock, or run a big-investor vote.

Name the ticker and make clear you want the panel's opinion on it. Make sure the cached files for that ticker already exist.

• 评审团怎么看这只股票？ (How does the panel view this stock?)
• 65 大佬怎么看 某某股票？ (What do the 65 big investors think of a given stock?)
• 某某会买吗？ (Would someone buy it?)
• 做一次大佬投票 (Run a big-investor vote)

What the output looks like and how to read it

Each persona returns a strict JSON object. The fields are investor_id, name, group, signal (bullish, neutral or bearish), confidence, score, verdict, reasoning, comment, pass, fail, ideal_price and period. The consensus is calculated as the share of bullish votes out of the total.

The comment is generated by the model in the investor's style. It is not a verified quote from that person. Treat the score and verdict as a structured prompt for discussion, not a forecast.

• signal: bullish, neutral or bearish
• confidence: a number from 0 to 100
• score and verdict: how strongly the persona endorses or rejects the stock
• comment: a model-written line in the persona's style
• panel_consensus: bullish votes divided by total votes, times 100

Safety and what the skill runs

The skill is marked as running scripts and as not using the network. The scripts it references are the Python modules listed above. Read those scripts before running the skill on your machine, since the package does not include all of them.

Because the personas are generated by the model, the skill can misrepresent real people. Do not present comments as things the investors actually said.

• Runs scripts: yes, according to the package metadata
• Uses network: no, according to the package metadata
• Licence: MIT

Tips for useful results

Give the agent complete cached files for the ticker. Missing dimensions usually mean personas cannot apply their rules, and the skill is meant to assign low confidence in that case.

Use the panel to compare philosophies. Look at where value, growth, macro and technical personas disagree, and ask why.

• Refresh the cache files before each run so the data is current
• Ask follow-up questions about a single persona's pass and fail lists
• Keep the MIT licence notice if you redistribute the skill