Causal inference design audit. Dispatches the Econometrician agent in standalone mode for 4-phase review (claim, design validity, inference, polish). Covers DiD, IV, RDD, Synthetic Control, and Event Studies. Use when working on empirical papers, strategy memos, or R scripts with causal estimators.
Run a 4-phase causal inference audit on the target file(s) by dispatching the Econometrician agent in standalone mode.
Determine target from $ARGUMENTS:
.tex file: Review paper for identification claims, assumption statements, estimation descriptionsstrategy-memo-*.md: Review strategy memo BEFORE code is written (early design check).R / .do / .py / .jl file: Review script for code-theory alignment, package usage, SE computationcode/analysis/): Review all scripts, then synthesizeoutput/paper/main.tex if it exists, else ask userBefore launching the reviewer:
Bibliography_base.bib to check citation availability.claude/rules/domain-profile.md for field-specific conventionsquality_reports/: read it for design intentDelegate to the econometrician agent via Task tool:
Prompt: Review [file] through all 4 phases of the econometrics review protocol.
Mode: Standalone (not within pipeline — skip orchestrator routing).
Focus on: [identified causal design if known, otherwise "identify the design first"].
Context: [brief summary of what the paper/script does, from Step 2].
Save report to: quality_reports/[FILENAME_WITHOUT_EXT]_econometrics_review.md
The agent runs 4 sequential phases:
Early stopping: If Phase 2 finds CRITICAL issues, the agent focuses there.
Present to the user:
Report the Econometrician's score. In standalone mode, this is the component score (not the weighted aggregate). Reference scoring-protocol.md for how this feeds into the overall score when run within the pipeline.