Contenu du pack

Un fichier SKILL.md (le guide pratique) + un modèle de départ. Aperçu :

Stock Researcher

When to use

• You want a deep-dive on a stock before investing — filings, earnings transcripts, real numbers, every claim cited
• You hold a portfolio and want a recurring research briefing (e.g. twice a week) on your holdings and watchlist
• You need the strongest counter-argument: a well-sourced bear case on a stock you already like
• You set risk rules (e.g. "flag me if it drops 15%") and want breaches surfaced automatically
• You're not a coder and want research notes in plain language with references you can actually check

Workflow

1. Set scope and sources (input: ticker + holdings/watchlist + risk rules → output: research brief)
   Give the AI the ticker (or a list of holdings plus a watchlist), your risk rules (stop-loss levels, max position size), and the sources it should trust: company filings first, earnings transcripts, then reputable financial press. You define the risk rules; the AI never sets or changes them.
2. Pull the primary documents (input: ticker → output: source bundle)
   The AI fetches the latest 10-K/10-Q, recent earnings call transcripts, and relevant news. Primary filings are evidence; press coverage is context. If a document can't be retrieved, the skill says so instead of guessing from summaries.
3. Extract and cite (input: source bundle → output: cited fact list)
   Summarize the business, the key numbers, and what management claims — with every claim tagged to its source and page. If a number can't be sourced, it's labeled unverified or dropped. One real case had the AI run the numbers twice to catch misreads.
4. Write the bear case (input: cited facts → output: bull/bear section)
   Deliberately argue the strongest case *against* the stock using the same cited facts. This is the skill's differentiator: a research note that tries to talk you out of the trade, not one that confirms what you already believe.
5. Save and spot-check (input: research note → output: saved note + verification pass)
   The note is saved as markdown in your notes library. Before acting on anything, you spot-check the citations — open the cited page, confirm the number. AI misreads tables; the citation exists so you can catch it.
6. Run the risk watch (input: schedule + risk rules → output: briefing with flags)
   On schedule, the AI refreshes the news, bull/bear arguments, and scenario analysis for your holdings and watchlist, and flags any position that broke your risk lines. Flagged doesn't mean sell — you still make every buy/sell decision yourself.

Output spec

Aperçu du modèle

Stock Researcher — Research Note Template

One note per ticker. Every number needs a source — write "unverified" rather than leaving it blank.

{Ticker} — {Company name} | {date}

Thesis question: {what you're trying to decide, e.g. "Is this a buy below $X?"} Sources used: {latest 10-K / 10-Q, earnings call {date}, news sources} Sources unavailable: {anything you tried to get but couldn't — "validate before relying"}

Business summary

{2–4 sentences: what the company does, how it makes money}

Key figures

Management claims vs. evidence

Bull case

1. {strongest pro, with citation}
2. {second pro, with citation}

Bear case

Le pack de compétence lui-même (SKILL.md + modèle) est rédigé en anglais.

Acheter le pack complet — $4.99 Pro — les 511 compétences