Read a Strange File Without Trusting It

What it is for

A structured reading request for code you did not write, demanding line citations and an explicit list of what the file cannot tell you. Use it on your first day in a repo, before touching a file nobody owns, or when tracing a bug through unfamiliar layers. It prevents confident summaries that quietly fill in error handling and side effects that are not actually there.

The skill Copy

Paste the file, or one function, and this brief. Do not ask "explain this
code" - you will get the code back as sentences.

Read this file and answer only these, in this order:

1. Entry points. Which functions are called from outside this file, and by
   what: a route, a job, a UI event, a test? Name them.
2. State. What does it read or write that outlives one call - database,
   cache, disk, globals, module-level variables? List each, and say who
   else can touch it.
3. Side effects. Every network call, email, payment, file write, or process
   spawn, with the line number.
4. The unhappy path. For each entry point: what happens on bad input, on a
   timeout, on a duplicate call? Quote the handling code, or say there is
   none.
5. Assumptions. What must be true elsewhere for this file to work - an
   order of calls, a field always present, a lock already held?
6. What you could not determine from this file alone, and which file you
   would need to see next.

Rules:
- Quote a line for every claim. No claim without a line.
- If something looks wrong, raise it as a question, not a fix. I have not
  asked for changes.
- Do not restate what the code obviously does. I can read it.

Failure this prevents: a fluent explanation you believe, that invented the
error handling and the transaction boundary you were about to rely on.

How to install it Copy

1. Open the tool and start a new conversation, project or template.
2. Paste the text below wherever the tool keeps standing instructions
   — usually called system prompt, instructions, or persona.
3. Save it so it applies to every run rather than a single message.

Did it work for you?

Where it came from