Claude, Earnings Preview Single.
The earnings-preview-single skill has Claude fill a fixed four to five page report template with earnings transcript, competitor, valuation, and news data, without ever forming an investment opinion of its own.
Claude picks up a skill called earnings-preview-single. Does that mean it starts analyzing the numbers and deciding whether a stock is worth buying? No. A skill is a folder Claude reads before it acts, and this one holds a report template plus an instruction file that tells Claude how to fill that template in, nothing more.
What sits inside the skill folder
Two files matter here. A report template, roughly four to five pages, lays out the structure of the finished document. A SKILL.md file tells Claude how to populate it: pull the latest earnings transcript, map the competitor landscape, note the valuation, and fold in recent news. That is the entire instruction set. There is no additional layer where Claude weighs the numbers and forms a view on whether the stock is a buy.
Proof by deletion
Delete the skills folder and Claude loses no investment opinion, because there was never one to lose. It simply stops filling in that particular template. That is the cleanest way to see what the skill actually is: a fixed structure plus a set of instructions for what to pour into it, not a repository of financial judgment.
Reading the file, then pouring inputs in order
Claude reads the SKILL.md and works through its steps in a fixed order: transcript, competitors, valuation, news, pouring each category into its place in the template shape. There is no branching unless the instruction file itself specifies one. That makes the process a template-filler, not a stock picker. The payoff is consistency: the same four to five page structure, filled from real inputs, every single run. The limit is exactly as narrow: the report never forms an opinion the template did not explicitly ask for.
What confidence and hedging in the output actually mean
A report that states a number with total confidence does not prove Claude checked that number against a second source. The template may simply call for confident language in that section regardless of how well-verified the underlying figure is. Conversely, a report that hedges a claim does not mean the underlying data was shaky. The template may ask for cautious phrasing in that section by design, independent of the actual data quality. Neither tone is evidence of verification one way or the other; the tone comes from the template's instructions, not from an assessment Claude made about its own confidence.
What the skill guarantees, and what it doesn't
A skill named earnings-preview-single never gives Claude a view on the stock itself. What it guarantees is that every run pulls the same categories of fact, earnings transcript, competitor landscape, valuation, and recent news, into the same report shape. Forming an actual investment opinion from that material remains entirely up to the person reading the report. Before relying on any such report, it is worth asking Claude to read the skill and walk through exactly what it will do before it does it, so you can see which categories of fact get pulled before the report gets written.
Key takeaways
- The earnings-preview-single skill fills a fixed report template; it does not analyze numbers to form an investment opinion.
- Two files define the whole process: a report template and a SKILL.md with the steps to populate it.
- Deleting the skill folder removes no investment judgment, because none existed in the first place.
- Confident or hedged language in the output reflects the template's instructions for that section, not evidence of how well the data was checked.
- The skill guarantees consistent structure and consistent categories of fact across every run, not an opinion on whether to buy.
Who this is for
Anyone using Claude to generate equity research or earnings preview documents who wants to understand exactly what the output represents, and what judgment still has to come from the reader.
Chapters
- 0:00The naive framing: "Claude decides if a stock is a buy"
- 0:10Sounds like analyst judgment
- 0:19Broken, with a case — delete the folder, nothing is lost
- 0:30The anchor: two files — report-template.md and SKILL.md
- 0:45Read it, then pour the inputs in order
- 0:56Template, not judgment — the payoff and the limit
- 1:07The anchor returns — same shape, every run
- 1:20Both directions — neither one is proof
- 1:37Carry-out
- 1:49Your turn
- 2:05Outro
Full transcript(auto-generated, with timestamps)
The naive framing: "Claude decides if a stock is a buy"
[0:00]You guess Claude decides whether a stock is a buy once it picks up the earnings preview single skill. It doesn't. It fills in a fixed report template. Let's see what's actually inside that file.
Sounds like analyst judgment
[0:10]Here, Claude has an earnings preview skill and it sounds like Claude is doing the analyst's job, reading the numbers and forming its own view on whether to buy, but nothing about its financial
Broken, with a case — delete the folder, nothing is lost
[0:21]Judgment changes. Delete the skills folder and Claude doesn't lose any investment opinion. There was none to begin with. It just stops filling in that one template. Here's what's
The anchor: two files — report-template.md and SKILL.md
[0:30]Actually inside. A license and two files that matter. A 44-page report template and a skill MD that tells Claude how to fill it in. Pull the latest earnings transcript. Map the competitor landscape. Note the valuation and fold in recent news. Claude reads that skill
Read it, then pour the inputs in order
[0:46]MD and works through the steps in order. Transcript competitors valuation news. Pouring each into the template shape. No branching unless the file itself says branch. That makes it a template filler,
Template, not judgment — the payoff and the limit
[0:58]Not a stock picker. The payoff, the same four to five page structure filled from real inputs every time. The limit, it never forms an opinion the template doesn't ask for. So, earnings preview
The anchor returns — same shape, every run
[1:08]Single never gave Claude a view on the stock. It just guarantees that every run pulls the same categories of fact, transcript, competitors, valuation, news into the same report template, MD shape.
Both directions — neither one is proof
[1:20]A report that states a number with total confidence doesn't prove Claude checked it against a second source. The template may simply present figures in a confident voice. In a report that hedges a claim doesn't mean the underlying data was shaky. The template may ask for cautious language in that section regardless. A skill named earnings
Carry-out
[1:38]Preview single doesn't hand claude a view on the stock. It's a template Claude fills in from the same categories of fact every time. And forming an actual investment opinion is still on you. Your turn. Paste this into Claude.
Your turn
[1:51]I want to generate a concise four to fivepage equity research earnings preview for a single company. Read the earnings preview single skill and walk me through what you will do before you do it. That way you'll see exactly which categories of fact it pulls before the report gets written. Clawed earnings
Outro
[2:06]Preview single, lay him in for bear.
More from Claude Basics
1:14Claude, Earnings Preview — What a Preview Skill Actually Previews
1:18Checks the Picture, Then Stops. — The FSI-Strip-Profile Skill (Anthropic Skills)
1:17Claude, Earnings Analysis — What One Skill File Actually Does
1:21Claude, Funding Digest — Does Claude Decide What's Newsworthy, or Get Told?
1:20Claude, Deal Tracker.
2:04