Skills 2026-03-23 06:37 UTC
stata-skillv1.1.0
Version [v1.1.0] committed by dylantmoore. Update Category: Feature Addition & Enhancement. Commit details: v1.1.0: Fix 4 doc bugs from eval, expand to 24 tasks, add contributor skill. This update introduces new capabilities and module extensions.
CommitRaw commit message
v1.1.0: Fix 4 doc bugs from eval, expand to 24 tasks, add contributor skill Eval results (24 tasks × 5 runs, Sonnet 4.6): Full skill 52.5/55 vs runner-only 37.9/55 (+14.6 delta) Skill wins 21/24, ties 2, loses 1 Bugs fixed (skill was actively teaching wrong patterns): - missing-data-handling.md: mi estimate without post before estimates store - missing-data-handling.md: tsset iter on panel-structured savetrace output - synth.md: ssc install synth_runner (not on SSC, needs net install) - binsreg.md: wrong savedata() variable names (bins_x vs actual dots_x) Gotchas added: - graphics.md: graph box incompatible with twoway - did.md: event-study dummy construction (hyphenated names, factor order) - sem-factor-analysis.md: nlcom clobbers e(), unstandardized for indirect effects - coefplot.md: standardized coefficient workflow Test infrastructure: - eval.py: plugins parameter for skill isolation, file-aware judge, --runner-only - 24 tasks (was 14), all rewritten as realistic research scenarios - runner-only-plugin for meaningful baseline comparison - Full eval results with run-by-run index in tests/EVAL_RESULTS_2026-03-22.md New: - stata-skill-contributor plugin: testing workflow, multi-agent analysis, cost guide - ROADMAP.md: gotchas-only variant, better test prompts, anchoring effect Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>