agentsclimarketplace
Sign in

Causal claims audit

Skill Nicowyn/paper-audit-skills/.agents/skills/causal-claims-audit

Audit causal claims in experimental and observational research. Use for randomized trials, difference-in-differences, event studies, instrumental variables, regression discontinuity, panel models, synthetic control, matching, mediation, and survey experiments when the task is to assess the estimand, assignment process, identifying assumptions, estimator, standard errors, robustness checks, and limits of interpretation.From its SKILL.md

Install
npx -y skills add Nicowyn/paper-audit-skills --skill causal-claims-audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 23 days oldThe repository was created 23 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.8 KB, 459 tokens by cl100k_base, as published. Nobody here has run it

Causal Claims Audit

Write the causal question and design before inspecting statistical significance.

Workflow

  1. State the target population, treatment, comparison, outcome, time window, and estimand.
  2. Reconstruct treatment assignment or the source of identifying variation.
  3. Write the assumptions that connect the observed comparison to the estimand.
  4. Identify the estimator, sample construction, fixed effects, controls, weights, and variance estimator.
  5. Check whether the reported diagnostics address each assumption and whether they have adequate power.
  6. Inspect treatment timing, spillovers, anticipation, attrition, interference, measurement, and post-treatment variables.
  7. Recompute at least one central estimate or diagnostic when artifacts permit.
  8. Separate:
    • a verified implementation or reporting error;
    • a violated or unsupported identifying assumption;
    • a robustness limitation;
    • an unresolved issue requiring data or design details.
  9. Calibrate the conclusion to the identified estimand and population.

Use references/design-checklists.md for method-specific checks and references/inference-and-diagnostics.md for standard errors, power, placebo tests, and diagnostic interpretation. Use $causal-inference-mixtape only for implementation templates, then verify that its template matches the paper’s design.

Rules

  • A pre-trend test that fails to reject does not prove parallel trends.
  • A first-stage F statistic above 10 does not prove instrument validity or universally rule out weak-instrument problems.
  • An overidentification test that fails to reject does not prove the exclusion restriction.
  • Covariate balance does not prove successful randomization when allocation concealment or attrition is compromised.
  • A smooth RD graph does not replace local identification and robust bias-corrected inference.
  • Conventional two-way fixed effects may mix invalid comparisons under staggered adoption with heterogeneous effects.
  • A robustness table cannot repair a mismatch between the estimand and the interpretation.

Output

Return the estimand, identifying variation, assumptions, evidence for and against each assumption, estimator and inference assessment, robustness gaps, external-validity limits, and a calibrated conclusion. Do not collapse these dimensions into one quality score.

What ships with it: 3 files

6.6 KB alongside SKILL.md

agents/

references/

Keep looking

Skills are one crate of 326,834. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.