Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add echoo19/hearth --skill hearth-playtestgit clone --depth 1 https://github.com/echoo19/hearthWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/echoo19/hearth/hearth-playtest)<a href="https://agentmods.dev/skills/echoo19/hearth/hearth-playtest"><img src="https://agentmods.dev/badge/skills/echoo19/hearth/hearth-playtest/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/echoo19/hearth/hearth-playtest"><img src="https://agentmods.dev/badge/skills/echoo19/hearth/hearth-playtest.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00112 | $0.02006 |
| Opus 5.5 | $0.00045 | $0.00802 |
| Sonnet 5.5 | $0.00022 | $0.00401 |
| Haiku 4.5 | $0.00011 | $0.00201 |
Grade A, and why
hearth-playtest scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bot playtesting: the engine plays your game
Scripted playtests (the hearth and hearth-feel skills) assert behavior you
already know to check. A sweep finds the behavior you didn't think to
check: bot policies play the scene headlessly across many seeds at full speed,
and hand you a compact report of softlocks, crashes, stuck states, and unreached
objectives. Because every run is deterministic, any failure is a perfect repro —
and you can freeze it into a permanent regression test. This is the closed loop:
play → judge → fix → guard.
Sweeps are read-only (no build permission). The bots produce evidence; you judge it — bots don't know if a game is fun, only whether it broke or stalled.
When to sweep
Three moments, non-negotiable:
- After any gameplay change. Zero setup —
mashworks on every game:
Let the engine try inputs you'd never think to try before you move on.hearth sweep "Level 1" --json - When a bug is reported — sweep for a repro FIRST, then debug. A failing seed is a deterministic, replayable repro; don't debug from a vague report. Find the seed, then open the systematic-debugging loop against it.
- Before calling a level "done." Prove the player can actually finish it:
seekthe exit with areachobjective (below). "It runs" is not "it's beatable."
Policies: which bot, when
Pass one or more with --policies (comma-separated). mash and idle need no
setup; wander and seek steer an avatar (the sole input-reading entity, or
--avatar <ref>), and seek also needs --target.
| Policy | What it does | Reach for it when |
|---|---|---|
mash |
Chaos monkey — random weighted-persistence input on every declared action. Zero config. | The default. Every sweep after a gameplay change; smoking out crashes and softlocks. |
idle |
No input at all. | Cutscenes, auto-play, timeouts — catch anything that breaks when the player does nothing. |
wander |
Curiosity-driven exploration; steers toward unvisited reachable cells. | Coverage — "is any part of this level unreachable or a trap?" |
seek |
Beelines an avatar to a fixed target. | Verification — "can the player actually get to the exit?" Pair with a reach objective. |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 149 lines · 112 tokens per session scan A 7dda20012f51
hearth-playtest is a skill published in the GitHub repository echoo19/hearth (49 stars, last pushed 19d ago), licensed MIT. It adds 112 tokens to every session and 2,006 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
smoke-check
Run the critical path smoke test gate before QA hand-off. Executes the automated test suite, verifies core functionality, and produces a PASS/FAIL report. Run after a sprint's stories are implemented and before manual QA begins. A failed smoke check means the build is not ready for QA.
gm-evaluate
Evaluate the current tag's quality: enforce the playable-closed-loop gate, maintain a single cross-tag e2e/ suite that always reflects the current game (add tests for new mechanics, prune tests for mechanics this tag deliberately removed), and reason about gameplay quality. Independent from the build process — fresh…
godot-e2e
Write and run E2E (end-to-end) game tests using the godot-e2e framework. Python controls a live Godot game over TCP — Locator-based semantic queries, expect() auto-retry assertions, and engine log capture make failures self-diagnosing. Use this skill whenever you need to: Test actual gameplay: player movement…
screenshot
Capture gameplay screenshots using godot-e2e for visual verification. Use when you need to: take a screenshot of the running game, capture multiple screenshots during an E2E scenario, generate reference.png, visually verify game state, or provide screenshots for VQA analysis. Triggers: "screenshot", "capture…
godot-testing-qa
Domain skill — Run a sequence of runtime actions and assertions. Assert runtime node existence and properties. Assert visible runtime text. Compare two PNGs using bounded pixel sampling. Sample runtime performance for a bounded frame count. Return the latest runtime test report. Gate collision acceptance on…
playtesting-a-feature
Use when about to claim a gameplay feature is done, shipped, or working — requires actually running the game and walking through the feature before declaring completion. Static diagnostics and type checks do not count as playtesting.