Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →brycewang-stanford/arpsych-comprehensiveness-and-balance
Use when checking that an Annual Review of Psychology (ARPsych) review covers the literature even-handedly — across labs, paradigms, and rival theories — weighs evidence by credibility, and avoids self-promotion. Audits coverage and fairness; it does not build the search (arpsych-literature-synthesis) or design the spine (arpsych-organizing-framework).
Passed 18 of 22 test cases with the skill, and 5 of 22 without it.
and 19 more in the full report
15 fixed3 passed either way2 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 15 test cases the skill fixed. Same model, same prompt, with and without the skill.
In a 2x2 taxonomy draft on attention types (focused vs divided, spatial vs feature-based), the author wrote 12 pages on spatial focused attention (their specialty) and one short paragraph on feature-based divided attention. Audit this manuscript using the balance format.
| Checked | Without skill | With skill |
|---|---|---|
| The response marks 【Cell depth】 as 'N' due to uneven coverage across taxonomy cells. | Missed | Passed |
| The response identifies deep coverage of favored topics versus thin coverage elsewhere as a balance failure. | Passed | Passed |
The instructions your model reads when the skill loads.
ARPsych authors are chosen partly for the standing to be even-handed, and the Committee assesses delivered reviews for accuracy, rigor, and balance (检索于 2026-06；以官网为准). A review of record cannot read as advocacy for the author's lab or favored theory. The job here is to audit the draft against four failure modes specific to high-stakes psychology reviews.
Use the evidence matrix's lab/camp and paradigm columns. If three groups or one method dominate the citations, ask whether that reflects the field or the author's reading. Actively seek the work you find least congenial — the strongest version of the opposing view, not a strawman.
Do not tally "12 studies found X, 4 found not-X." Weigh by design strength, sample, and replication status. A single large, preregistered, multi-site study can outweigh a dozen small underpowered ones. Post-replication-crisis, ARPsych readers expect you to flag effects that failed to replicate rather than reviewing them as established (检索于 2026-06；以官网为准).
| Signal | Up-weight | Down-weight |
|---|---|---|
| Replication | direct + conceptual replications | single-lab, never replicated |
| Power/sample | large, well-powered, diverse | small, WEIRD-only, p≈.05 |
| Design | preregistered, controlled, longitudinal | post-hoc, cross-sectional, flexible analysis |
| Effect realism | meta-analytic, corrected for bias | uncorrected, funnel-asymmetric |
Present rival accounts at their strongest, then adjudicate on evidence — do not settle a live debate by authorial fiat. If the field is genuinely unresolved, say so; an honest "we don't yet know" is more authoritative than a forced verdict.
Count citations to the author team versus comparable groups. The author's work belongs in the review in proportion to its place in the field, not as the center of gravity. Reviewers and the Committee notice citation self-concentration.
test cases passed, vs 5 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.