Start from the job in front of you. Each workflow states what it is for, what it needs, what it produces, what you check, where its authority stops, and which file it runs on. One file can serve more than one job.
The investment process
The article uses five broad stages. In the Toolkit, I split the first stage into two separate jobs: 1A. Idea Generation, which fills the funnel, and 1B. Idea Triage, which determines what deserves deeper research. The workflows are organized by the job you are doing, not by the underlying asset. When one asset supports more than one job, it appears in more than one stage. Use the Table of Contents at the top of this page to jump to the stage you need.
What to expect from the Claude Skills
Claude Skills are useful but imperfect. Current models can miss a fact, misread a source, skip a field you asked for, use something stale, or state a conclusion with more confidence than the evidence carries. Those limits are structural. They vary by model and they vary by run.
Even with those limitations, I have found the workflows I am sharing with you to be very useful as long as they are used for their intended purpose. They save time and help you find patterns that you might have missed. As a result, I have been able to look at a lot more investing candidates and triage them much faster than I could have done without the help of AI.
Every Claude workflow below was run on a representative case before it was listed here. Its card carries what you have to verify, and its guide records what I saw go wrong. The intended output is a target. It is not a promise of identical format, identical coverage, or an identical conclusion across models and runs.
Files live in Downloads → available September 8. Downloads is a web page, so open it in a browser rather than in the Substack app.
Before you start
For these Claude workflows, use Claude at claude.ai in a web browser. Install and enable the relevant skill, and turn on Code execution and file creation.
To use the full collection, you will need a separate paid Claude subscription, such as Pro or Max. For workflows marked Research required, turn on Research and web search before starting.
The Mental Model Scorecard and PULSE do not require Claude.
1A. Idea Generation
Special Situations Country Sweep
Use it when: you want a country-level list of recent and upcoming event-driven situations before deciding where deeper work is worth the time.
Needs: a country, a lookback and look-forward window, and an optional market-cap floor.
Claude setup: paid Claude plan · Research required · web search enabled
Produces: a sourced triage list with category counts, catalyst timing, and a dated methodology note.
You check: a sample of the tickers, the cited source for anything you pursue, and whether the event is still live.
Where it stops: a name on the list is a lead. It is not evidence that a security is mispriced or that it deserves deeper work.
Runs on: Special Situations Screener · Access: all paid
The Canada runs were useful for finding leads and they were not clean. Coverage was incomplete, some filters and tickers came back wrong, a few statuses were stale, fields went missing, duplicates appeared, and some lines read closer to a recommendation than the evidence justified. Treat the output as a starting list to verify, never as a screen result you can act on.
1B. Idea Triage
Build a Sourced Company Overview
Use it when: you need a structured first orientation to a public company and its industry.
Runs on: Company Overview Deck · Access: free
See how it works: Read the article
Try it free: Get the Company Overview Deck Skill and use code YZMDK4NQ for 100% off. No payment information is required after you apply the code.
Score and Prioritize a New Idea
Use it when: you want a structured score on a new idea, including a written bear case, before you decide it is worth real work.
Access: limited browser access for everyone; the full Mental Model Scorecard for paid subscribers saves, compares, and analyzes entries
Runs on: Mental Model Scorecard
Route: Open the Scorecard
Run a Fast PULSE Triage Pass
Use it when: you want a fast read on economic profitability, free-cash-flow behavior, leverage, and valuation before choosing what deserves deeper work.
Needs: one equity ticker and the Full PULSE template with Bloomberg codes, opened in desktop Excel with the Bloomberg Excel add-in installed and connected to an active Bloomberg subscription.
Produces: a refreshed PULSE dashboard, the charts, the historical data, and a list of questions to follow up.
You check: Make sure Bloomberg pulled the right company, currency, periods, and current balance sheet. Then look for special charges or other one-off items that could distort the calculations. Bloomberg data is useful, but it is not perfect.
Where it stops: PULSE does not normalize accounting, settle a valuation, or produce a recommendation.
Runs on: Full PULSE template with Bloomberg codes · Access: annual and founding
Test a New Idea Before Committing to Deep Research
Use it when: a ticker has reached your short list and you want to understand the strongest case against it before committing substantial research time.
Needs: one public company.
Claude setup: paid Claude plan · Research required · web search enabled
Produces: an independent bear-case memo, a stress-test checklist, stated source limitations, and numbered sources.
You check: the latest filings, current balance-sheet and valuation support, transcript citations, and the key claims the argument depends on.
Where it stops: this is an early warning pass. It does not decide whether the company deserves deeper work.
Runs on: Devil’s Advocate · Access: all paid
2. Research
Map the Competitive Set
Use it when: you are starting deep research and need to decide whose filings, transcripts, industry material, and briefing sources belong in the set.
Needs: one public-company ticker.
Produces: a sourced list of direct competitors and substitutes.
You check: the latest annual filing, segment coverage, current ownership and listings, and legal domicile.
Where it stops: this is a research set, not a valuation peer group. Appearing on the list does not mean the economics are comparable.
Runs on: Competitors · Access: all paid
The Accenture run held up with caveats. Verify the ordering, the competitor-versus-adjacent classification, ownership, listing, domicile, and the overlap logic before you let the list decide where your reading time goes.
Management Compensation and Incentives Review
Use it when: you want to know what actually drives the CEO’s incentive payout, not what the summary paragraph says it drives.
Needs: one U.S. public-company ticker for a domestic issuer that files a DEF 14A.
Produces: the filing metadata, the incentive metrics and their weights, CEO compensation, and an ROI Influence score or a specific reason it cannot be computed.
You check: that it used the definitive proxy, the fiscal year, any CEO transition, the metric classifications, every reported weight, and whether each percentage describes a performance level or a payout level.
Where it stops: the score describes the payout structure. It says nothing about capability, honesty, shareholder orientation, or whether the investment is any good.
Runs on: Proxy Analysis · Access: all paid
The Home Depot run reached the right conclusion and got there with real gaps. Long-document retrieval dropped disclosed short-term weights and the disclosed compensation figure, labeled a disclosed pay mix as an assumption, computed the score from a default after flagging a missing input, and confused performance levels with payout levels. Read the proxy sections it cites.
3. Investment Decision
Build the Strongest Bear Case Before Acting
Use it when: a case is complete enough that it deserves an independent, sourced attempt to break it.
Needs: one public company.
Claude setup: paid Claude plan · Research required · web search enabled
Produces: a bear-case memo with evidence tags, a stress-test checklist, stated source limitations, and numbered sources. Where the source coverage is too thin, it should produce a preliminary checklist instead of a confident memo.
You check: the latest annual and operating-period filings, current balance-sheet and valuation support, transcript citations, and at least 5 of the claims the argument rests on.
Where it stops: it does not recommend a trade, a price target, or a position change. Weak evidence should produce a weak or mixed bear conclusion, and if it produces a strong one anyway, that is the thing to check first.
Runs on: Devil’s Advocate · Access: all paid
5. Monitoring and Information Processing
Re-score a Position After a Material Event
Use it when: earnings, a filing, or a company event has landed and you want to know whether your view moved or only your mood did.
Score the same company again in the Scorecard and put the two entries side by side, question by question. The direction and size of the change are the signal. A saved entry cannot be edited after the fact, which is the point.
Runs on: Mental Model Scorecard · Access: all paid
Route: Open the Scorecard
Revisit an Existing Holding With a Fresh Bear Case
Use it when: you have followed or owned a company long enough that prior assumptions and behavioral biases may be shaping what you see.
Needs: one public company.
Claude setup: paid Claude plan · Research required · web search enabled
Produces: an independent bear-case memo, a stress-test checklist, stated source limitations, and numbered sources.
You check: the latest filings, current balance-sheet and valuation support, transcript citations, and the key claims the argument depends on.
Where it stops: it can identify what deserves another look. It does not tell you whether to change a position.
Runs on: Devil’s Advocate · Access: all paid
Process Review
Investment Autopsy Library, 22 case studies
The free Season 1 recap is the index. It groups all 22 cases by the lesson that repeated and carries the full case appendix. Individual cases keep the access they already have.
Access: free index
Route: Browse the library
For members who joined before September 8
Short version: your access grew, and one file may need replacing.
What is new to you at the flip: the Downloads-backed workflows above, the Downloads library behind them, and the full Mental Model Scorecard. Annual and founding members also get the Full PULSE template with Bloomberg codes and office hours.
What you already have is still yours. All 22 Investment Autopsy cases, indexed by lesson, at the link above.
Whether to download again: if you bought a skill file from me before September 8, download the current version from Downloads. Several packages have changed since their original release, and the Downloads copy is the one the workflow guides describe.
Where access lives now: Downloads for the files, The Toolkit for the workflow that uses each file. Downloads is a web page, so use a browser rather than the Substack app.
What ended: the weekly Investment Autopsy finished as a completed 22-case library in July, and the Subscriber PULSE Check is retired. Both are covered in the member note I sent on September 4.
Questions about any of this, reply to that note and it comes straight to me.
The boundary, stated plainly
These workflows organize research and challenge assumptions. The evidence, the valuation, and the decision stay yours.
Full terms: [Membership boundaries] → https://newsletter.behavioralvalueinvestor.com/p/membership-boundaries