How do I install the Playtests skill?

Run npx skillmds@latest add lambdahack/playtests 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 Playtests safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Playtests?

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 Playtests free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Playtests?

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

Playtests by lambdahack — Agent Skill

Name: Playtests
Author: lambdahack

Playtests

LambdaHack's automated AI-vs-AI playtest and benchmark Makefile targets (test-short, test-medium, test-gha, frontend*, bench*) and the headless test-frontend flags. Use before running, interpreting or adding a playtest or benchmark target.

lambdahack 42ca237 2.2 KB Updated Sep 17, 2026

File contents

Playtests and benchmarks (Makefile)

The names cited here --- Makefile targets, frontend flags, document paths --- were verified against the tree at commit ef4d44732 (2026-07-30); the reference pass proves a cited name exists, this stamp that it still does what the claim around it needs. The wording differs from the stamp in CLAUDE.md and test/CLAUDE.md because this file cites no line numbers.

The Makefile has a large battery of automated AI-vs-AI playtest and benchmark targets: make test-short, make test-medium, make test (those two plus benchNull), make test-gha (a larger aggregate plus test-sniff; run by CI on each push), make frontendCrawl/make frontendBattle etc. (interactive AI-vs-AI games in the SDL2 frontend, useful to visually confirm a change), and make bench* targets for performance. test<Mode>-medium targets run one game mode each -- testRaid-medium, testBrawl-medium, testShootout-medium, testHunt-medium, testFlight-medium, testZoo-medium, testAmbush-medium, testCrawl-medium, testSafari-medium, testBattle-medium, testDefense-medium, testDig-medium and a few variants -- through the teletype frontend with --automateAll. Note the capitalised mode and the single hyphen: there is no test-raid-medium. Grep the Makefile for a mode name to find its exact invocation before adding a new one.

The headless targets select test frontends by flag: --frontendNull (frames forced but not displayed), --frontendLazy (frames not even computed), --frontendTeletype (line-printer output). The nodeBench* and nodeMinifiedBench targets are dead GHCJS remnants --- they invoke a .jsexe that nothing builds anymore; repurposing them for WASM is Phase 3 of docs/wasm-frontend-unified-plan.md.

Such runs look completely silent, because the main game executable redirects its own stdout and stderr to files --- see "Playtests and headless runs" in the repo-root CLAUDE.md for how to harvest the output.

---name: playtestsdescription: LambdaHack's automated AI-vs-AI playtest and benchmark Makefile targets (test-short, test-medium, test-gha, frontend*, bench*) and the headless test-frontend flags. Use before running, interpreting or adding a playtest or benchmark target.---# Playtests and benchmarks (Makefile)The names cited here --- Makefile targets, frontend flags, document paths ---were verified against the tree at commit `ef4d44732` (2026-07-30); the referencepass proves a cited name exists, this stamp that it still does what the claimaround it needs. The wording differs from the stamp in `CLAUDE.md`and `test/CLAUDE.md` because this file cites no line numbers.The Makefile has a large battery of automated AI-vs-AI playtest and benchmarktargets: `make test-short`, `make test-medium`, `make test` (those two plus`benchNull`), `make test-gha` (a larger aggregate plus `test-sniff`; run by CIon each push), `make frontendCrawl`/`make frontendBattle` etc. (interactiveAI-vs-AI games in the SDL2 frontend, useful to visually confirm a change),and `make bench*` targets for performance. `test<Mode>-medium` targets run onegame mode each -- `testRaid-medium`, `testBrawl-medium`, `testShootout-medium`,`testHunt-medium`, `testFlight-medium`, `testZoo-medium`, `testAmbush-medium`,`testCrawl-medium`, `testSafari-medium`, `testBattle-medium`,`testDefense-medium`, `testDig-medium` and a few variants -- throughthe teletype frontend with `--automateAll`. Note the capitalised modeand the single hyphen: there is no `test-raid-medium`. Grep the Makefilefor a mode name to find its exact invocation before adding a new one.The headless targets select test frontends by flag: `--frontendNull` (framesforced but not displayed), `--frontendLazy` (frames not even computed),`--frontendTeletype` (line-printer output). The `nodeBench*`and `nodeMinifiedBench` targets are dead GHCJS remnants --- they invokea `.jsexe` that nothing builds anymore; repurposing them for WASM is Phase 3of `docs/wasm-frontend-unified-plan.md`.Such runs look completely silent, because the main game executable redirectsits own stdout and stderr to files --- see "Playtests and headless runs"in the repo-root `CLAUDE.md` for how to harvest the output.

Frequently asked questions