Compare Fictional Equity Theses With the Agentii Long Short Skill

Use Agentii’s publisher-maintained Claude Code research framework to compare expectation gaps, document competing evidence, and set falsification checks.

Setup for Claude Code

Claude Code and a scratch Git project. The registered skill requires an Agentii API key and its HTTPS MCP endpoint for ticker resolution, coverage and historical evidence. The worksheet below exercises the published method locally without an account or live market data.

Clone the publisher's repository with `git clone https://github.com/agentii-ai/agentii-investment-intelligence.git`. In the checkout, run `git checkout --detach 9b58878b79866528153dc108902aef58b571eb78`.
From the pinned checkout, run `bash scripts/copy-skills-local.sh /absolute/path/to/an/empty-practice-project`. The publisher's local installer stages command wrappers and full skill folders with their references.
For connected research in Claude Code, follow the pinned `QUICKSTART.md` to configure Agentii MCP and set `AGENTII_API_KEY` outside the project. Restart Claude Code in the practice project and check whether `/agentii:long-short-equity` appears in the command menu.

Example task

In a scratch file, use two invented businesses and next-year operating profit in the same units. Reefline has an assumed expectation of 100 and a research case of 115. Pineharbor has an assumed expectation of 80 and a research case of 68. Ask Claude Code to use the pinned long-short skill document as a reference for an offline worksheet. Calculate (research case − expectation) / expectation for each, label the input assumptions, identify one fact that would overturn each research case, and flag missing historical and execution inputs.

Expected result

The worksheet shows a +15% gap for Reefline and a −15% gap for Pineharbor. It records two falsification checks and leaves historical precedents, real market positioning and execution status open. The exercise ends with a comparison of assumptions and evidence.

Upstream documentation

Original SKILL.md on GitHub · Apache-2.0 license · Source reviewed 2026-11-15

Frequently asked question

Does the example verify the connected Claude Code command?

The fictional example verifies the comparison arithmetic and how to document missing evidence. The registered command also needs successful authenticated ticker and coverage checks through the publisher's MCP connector.