A Literature Scan Someone Can Repeat

What it is for

A working method for building a defensible reference list: split the question, log every query with its filters and date, search claims in both directions, sort by study type, and verify every citation by opening it. Reach for it when the bibliography will be reviewed, graded or relied on by someone else. It prevents a list nobody can reproduce and a citation that does not exist as written.

The skill Copy

Use this when the output is a reference list other people will rely on.

1. Write the question down before searching, and split it into three or four
   sub-questions. Search each separately. One broad search returns the popular
   papers, not the relevant ones.
2. Keep a log as you go, one row per search: the exact query, the filters, the
   date, and how many results you kept. Without it nobody - including you next
   month - can tell whether a gap in the list is a gap in the evidence or a
   query you never ran.
3. For each sub-question, search the claim and its negation. Stop adding
   papers when three searches in a row return nothing new.
4. Sort what you keep by study type before reading: reviews and meta-analyses
   first, then trials, then observational work, then everything else. The top
   layer tells you which of the lower layers are worth your afternoon.
5. Treat this as a starting point, not a complete search. An abstract-level
   tool misses findings that appear only in a results table, and misses papers
   it has no record of. When the stakes are high, repeat the key searches in a
   database that indexes your field properly.
6. Record for each kept paper: full citation, year, design, sample, the
   finding in one line, and the section you took it from. Do this while the
   paper is open, never later from memory.
7. Verify every citation exists exactly as written - title, authors, year,
   journal, identifier - by opening it. Check the whole list, including the
   ones you are sure about.
8. Write down what you did not find. A sub-question with no usable evidence is
   a result, and it belongs in the write-up.

The failure this prevents: a bibliography that looks thorough, cannot be
reproduced, and contains one entry that does not say what the list claims.

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