How do I install the Testing skill?

Run npx skillmds@latest add nethercore-systems/testing in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Testing safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Testing?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Testing free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under Apache-2.

Who published Testing?

nethercore-systems (@nethercore-systems) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Testing by nethercore-systems — Agent Skill

Name: Testing
Rating: 4.2 (5 reviews)
Author: nethercore-systems

File contents

Nethercore Testing

Real replay regression

From a game project with nether.toml and an authored tests/smoke.ncrs:

nether build
nether replay validate tests/smoke.ncrs
nether replay run tests/smoke.ncrs --rom game.nczx --headless --report replay-report.json --timeout 30

Replace game.nczx with the ROM produced from [game].id. The matching nethercore-zx executable must be alongside nether or discoverable by its player lookup. Validation checks syntax only; only ROM-backed execution tests actual gameplay. No recording command is provided: author .ncrs text scripts.

Require both a zero exit status and a completed PASSED report. A timeout, missing action/variable, trap, failed assertion, or desync is not a pass. Keep an intentionally false assertion as a negative control when establishing CI. Use fresh report paths; never interpret an old report after a failed invocation.

Expose meaningful state

Register debug_watch_* or debug_register_* pointers in init() for stable, rollback-backed fields: phase, health, position, score, current rule state. The runner reads these through the existing debug registry. Use group-qualified names shown in the report, not guessed names or fixed WASM addresses. Keep pointers valid; for movable game structs use stable shadow statics refreshed in update() and after debug actions. Actions set up scenarios; inputs must still test the actual rules. Discover available names with a snap = true frame before writing assertions.

See references/replay-format.md for a complete discovery script and frame semantics. Check sound, readability, and game feel through rendered playback and human review separately; headless success is not an artistic verdict.

Rollback determinism

nether run --sync-test --exit-after-frames 1000

This uses GGRS to restore/re-simulate and compare checksums; it is not two independent rendered game instances. The command above requires a display. A bounded successful run proves only that tested sequence, not every game path or cross-platform determinism. Test input-rich game scenarios as well as idle.

State belongs in WASM linear memory plus the engine's rollback state. Mutable WASM globals/tables are not snapshotted; avoid memory growth after init. Use engine RNG/time/input and keep simulation mutations out of render().

Executable cross-game check

From this plugin repository, with the sibling engine binaries and game ROMs built:

python scripts/verify_agent_workflow.py

This runs semantic scenarios for Volley Fighter and Scrapheap Saints, repeats them to compare state snapshots, and checks deliberate failures. Missing game ROMs fail rather than skip. Pass explicit paths with --help for other layouts.

---name: testingdescription: Test Nethercore games with real ROM-backed replay assertions, semantic state snapshots, debug actions, and bounded rollback sync tests. Use for regression testing, automated game verification, and desync diagnosis.license: Apache-2.0---# Nethercore Testing## Real replay regressionFrom a game project with `nether.toml` and an authored `tests/smoke.ncrs`:```bashnether buildnether replay validate tests/smoke.ncrsnether replay run tests/smoke.ncrs --rom game.nczx --headless --report replay-report.json --timeout 30```Replace `game.nczx` with the ROM produced from `[game].id`. The matching`nethercore-zx` executable must be alongside `nether` or discoverable by itsplayer lookup. Validation checks syntax only; **only ROM-backed execution testsactual gameplay**. No recording command is provided: author `.ncrs` text scripts.Require both a zero exit status and a completed `PASSED` report. A timeout,missing action/variable, trap, failed assertion, or desync is not a pass. Keepan intentionally false assertion as a negative control when establishing CI.Use fresh report paths; never interpret an old report after a failed invocation.## Expose meaningful stateRegister `debug_watch_*` or `debug_register_*` pointers in `init()` for stable,rollback-backed fields: phase, health, position, score, current rule state.The runner reads these through the existing debug registry. Use group-qualifiednames shown in the report, not guessed names or fixed WASM addresses. Keeppointers valid; for movable game structs use stable shadow statics refreshedin `update()` and after debug actions. Actions set up scenarios; inputs muststill test the actual rules. Discover available names with a `snap = true`frame before writing assertions.See `references/replay-format.md` for a complete discovery script and framesemantics. Check sound, readability, and game feel through rendered playbackand human review separately; headless success is not an artistic verdict.## Rollback determinism```bashnether run --sync-test --exit-after-frames 1000```This uses GGRS to restore/re-simulate and compare checksums; it is not twoindependent rendered game instances. The command above requires a display.A bounded successful run proves only that tested sequence, not every game pathor cross-platform determinism. Test input-rich game scenarios as well as idle.State belongs in WASM linear memory plus the engine's rollback state. MutableWASM globals/tables are not snapshotted; avoid memory growth after init. Useengine RNG/time/input and keep simulation mutations out of `render()`.## Executable cross-game checkFrom this plugin repository, with the sibling engine binaries and game ROMs built:```bashpython scripts/verify_agent_workflow.py```This runs semantic scenarios for Volley Fighter and Scrapheap Saints, repeatsthem to compare state snapshots, and checks deliberate failures. Missing gameROMs fail rather than skip. Pass explicit paths with `--help` for other layouts.

Frequently asked questions

Testing