| name | regression-surface |
| description | Run the correct AlphaGalerkin regression-surface test command block for a changed code path. Use after editing solver/evaluator/PDE/scenario/agent/codec code to run exactly the guarding test suites the CLAUDE.md Regression Surface table prescribes, instead of guessing which tests to run. |
regression-surface — run the guarding tests for a change
CLAUDE.md maintains a Regression Surface table mapping each subsystem to the exact test
command(s) that must stay green when it changes. This skill selects and runs the right block.
Steps
- Identify the changed surface from the edited files, then open the Regression Surface
table in
CLAUDE.md and find the matching row(s). Key mappings:
src/alphagalerkin/** → Solver wiring + Trained evaluator + per-module coverage rows.
src/pde/**, src/poc/scenarios/_centaur_common.py → PDE e2e + LLM-prior + scaling-law +
research-loop rows (the shared centaur primitives fan out to all three).
src/poc/scenarios/noyron_basis* , src/pde/operators_picogk.py, src/pde/sdf.py,
src/pde/geometry_picogk.py → Noyron HX scenario row.
src/integrations/** → LLM-prior (mocked CPU) + backend-registry rows.
src/agents/** → Centaur research-loop + agents coverage rows.
src/video_compression/** → the relevant Phase 0/1/2 codec rows (separate coverage workflow).
- Run the block with
-m "not gpu_required" unless CUDA is available; the root conftest.py
auto-skips GPU tests on CPU hosts.
- Run the coverage gate row if the change is non-trivial (
--cov-fail-under per module).
- Report pass/skip/fail faithfully — never claim green on a suite you did not run.
Notes
- If the change adds a new surface, add a new row to the table (see the
spec-new skill).
- The dependency notes under the table list cross-surface fan-out — honor them.