Menú
Koru regression test harness workflow. Use when checking test status, investigating failures, or coordinating test runs with the user.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Tend the membrane memory corpus — an OKF concept store whose fragment evolution lives in the git log as lineage trailers. Fire when creating, evolving, merging, splitting, or CORRECTING durable memory/knowledge that lives as OKF files under git; when recording how a belief changed or was repudiated; or when tracing a concept's history through time. YOU are the evolver — there is no evolution engine and no LLM sidecar. NOT for transient notes, scratch, or code comments.
Compile-time metaprogramming in Koru: event globbing and compile-time event matching — the "you are the compiler" surface where userland code sits inside the pipeline. Use when writing compile-time event selection or pattern-matching over events. SEED TOPIC — currently scoped to event globbing; widen `clusters` in koru-by-example.json as the metaprogramming surface grows.
Control flow in Koru: `~if` and `~for` as templates over continuation-return, not host keywords. Use when emitting conditionals or iteration in Koru, or when working on the template/lowering layer (`koru_std/control.kz`). Covers how `for`/`if` select continuations and lower per target.
How to work with the Koru language and its compiler `koruc`. Use when working in a Koru repo (`.kz`/`.k` files), running or reading regression tests, or when you need to orient on the toolchain — what the language is called, how to compile, where the docs and tests live. Start here before reading code.
How to work with the Koru language and its compiler `koruc`. Use when working in a Koru repo (`.kz`/`.k` files), running or reading regression tests, or when you need to orient on the toolchain — what the language is called, how to compile, where the docs and tests live. Start here before reading code.
Drive koru's diagnostic-code registry to green. Use when `zig run scripts/registry_check.zig` FIRES on DEAD/ORPHAN_EMIT/ROTTEN_PIN codes and you need to disposition each one (wire an emit / reserve / consolidate). The watcher detects drift in src/errors.zig's ErrorCode enum; this is the playbook for clearing it. Derived from three cold-room probes — trust the procedure, verify every claim.
| name | regression |
| description | Koru regression test harness workflow. Use when checking test status, investigating failures, or coordinating test runs with the user. |
The regression suite is the hub for all Koru compiler progress. ~600 tests, 40+ minutes for a full run.
The user runs the full suite (it takes too long for me to run):
./run_regression.sh # Full run
./run_regression.sh --parallel 8 # Parallel (faster)
I check what broke and investigate:
./run_regression.sh --regressions # What's failing + when it last passed
./run_regression.sh --status # Current state from disk markers
./run_regression.sh --diff # Compare current vs last snapshot
We debug specific tests together:
./run_regression.sh 501 # Run single test (by number)
./run_regression.sh 330 # Run range 330-339
./run_regression.sh --history 501 # When did this test break?
After fixing, verify:
./run_regression.sh 501 502 503 # Re-run affected tests
./run_regression.sh smoke # Quick sanity check
| Command | Purpose |
|---|---|
--status | Show current test state from disk markers |
--regressions | List failing tests + when they last passed |
--diff | Compare current run vs last snapshot |
--history <id> | Show test history across all snapshots |
--list | List all tests with descriptions |
--priority | Show tests marked as PRIORITY |
--last-run | Show results from last full run |
Tests are marked with files in their directory:
SUCCESS / FAILURE - Result of last runTODO - Not yet implementedSKIP - Intentionally skippedBROKEN - Known broken, needs fixBENCHMARK - Performance test, not run normallyPRIORITY - Urgent, needs attentionAfter each full run, a snapshot is saved to test-results/ with:
--diff and --history--regressions first when the user reports failures.