Use when positioning a manuscript against the existing economics literature, building the antecedents map for the introduction, deciding what to cite, or verifying that every reference in the bibliography is real, correctly attributed, and cited to the published version. Apply at topic selection for the novelty scan and again before drafting the introduction.
SkillScore82Maintained
Automatic
◆Research pack
Last verified 8/8/2026
What it changes
Passed 18 of 22 test cases with the skill, and 10 of 22 without it.
Without skill45%
With skill82%
Checks it passed that the model alone missed
Identifies searching by dataset name as a specific search tactic
Structures the antecedents matrix using the column header What it does not do
Replaces the working paper reference with the published journal version reference
and 9 more in the full report
10 fixed8 passed either way2 still failing2 broke
Pass-rate lift
+36 pp
Tokens
+90%
Agent turns
0%
Response time
-15%
Model gemini-3.6-flashMeasured 8/8/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
Here are three raw bib entries I typed up for my paper: `@article{CardKrueger1994...}`, `@techreport{Acemoglu2001...}`, `@article{autor-duglo-2003...}`. Please format these BibTeX keys into the standard style used by top economics working paper repositories.
What was checked, without and with the skill
Checked
Without skill
With skill
Re-formats all BibTeX keys into lowercase author_author_year style without camelCase or uppercase letters
Missed
Passed
case-05 · gemini-3.6-flash · 8/8/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md219 lines2 files
AER Literature
Overview
This skill does two jobs that AER referees grade separately:
Positioning — finding the closest predecessors, mapping what each did
and missed, and choosing the value-added claim the paper can actually
defend. A mispositioned paper draws the fatal report sentence "the authors
appear unaware of [paper]."
Citation integrity — guaranteeing that every reference exists, says
what the manuscript claims it says, and is cited to the right version.
The second job is non-negotiable for AI-assisted manuscripts. Language
models produce plausible-looking citations that do not exist, attach real
authors to invented titles, and attribute findings to papers that show the
opposite. One hallucinated reference, found by a referee, destroys the
credibility of every other sentence in the paper. The working rule:
No citation is written from memory. Every entry is verified against a
fetched source before it enters the bibliography, and every claim about a
paper is verified against that paper's actual text.
When to Use
During aer-topic-selection, to run the novelty scan before committing
Before aer-introduction, to build the antecedents paragraph
When a referee names a missed paper and the positioning needs repair
Before submission, as the bibliography verification pass (pairs with
aer-consistency)
Search Protocol
Run all five channels — each finds papers the others miss:
Working-paper series: NBER, CEPR, IZA, SSRN, BFI. Top-5 competitors
live here for 2-4 years before publication; the novelty threat is almost
always an unpublished working paper, not a published article.
RePEc / IDEAS and Google Scholar, searching by question ("effect of X
on Y"), by method-in-setting, and by dataset name. Dataset search catches