Skip to content
Get packs

Investment Thesis Drift Checker

Compares two dated investment thesis reports and flags only evidence-based changes, not rewording

Source Download skill
Repository
xbtlin/ai-berkshire
Stars
16,982
Licence
MIT
Files
1 file

What it does

This skill teaches an AI agent to compare an older and newer investment research report on a company and decide whether the underlying investment case has actually changed. It separates real factual changes (revenue, margins, governance, competitive position, red-flag events) from price moves and from mere wording differences, and it asks the agent to cite specific evidence for every change it reports. It is written in Chinese and is a Codex adapter of a longer workflow in the AI Berkshire project.

Great for

  • Long-term stock investors who keep written thesis reports and want a disciplined review after new earnings or major news
  • Analysts who maintain dated research snapshots and need a structured comparison table
  • Teams building a repeatable process to avoid reacting to share price swings or rephrased reports

Skip it if

  • Anyone wanting buy or sell advice without their own research: the skill produces analysis built on the user's supplied reports, and its outputs depend heavily on the quality and completeness of those files
  • Users without an existing thesis baseline: the skill expects a prior structured report (ideally created by a companion thesis-tracker skill) and will refuse to run drift detection without one
  • Non-Chinese speakers who need English output: the instructions are written in Chinese and reference Chinese-market filings

Try asking

  • “Compare reports/Apple-thesis-2024-01.md with reports/Apple-thesis-2025-01.md and tell me whether the investment thesis drifted”
  • “Run thesis drift detection for 贵州茅台 using the latest two dated snapshots in the reports folder”
  • “Check whether my moat assessment for this company changed after its latest annual report”
Step-by-step guide: how to use Investment Thesis Drift CheckerInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    thesis-drift.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip thesis-drift.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip thesis-drift.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Provide two dated report paths explicitly to avoid ambiguous baseline selection
  • Make sure your baseline report contains explicit hypotheses, red lines and valuation anchors, otherwise the skill will mark many dimensions as unable to judge
  • Verify any numbers the agent reports against the original filings, since the skill itself does not fetch data and relies on figures already in your reports

What's inside

  • SKILL.md

Redistributed unmodified from xbtlin/ai-berkshire under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.