- Home
- Skills
- Testing & Debugging
- autoplay balance test
autoplay balance test
Headless playtesting bot for browser games that identifies difficulty walls, level curve spikes, and console errors.
Free
Works with the AI Agents you already use
autoplay balance test
Example session with this skill installed
Run a balance test on our new dungeon crawler. The adapter is at qa/dungeon-adapter.mjs. I need to know if the level 5 boss is too hard.
- Read your context and instructions
- Compiled the autoplay balance test
I have completed the balance analysis for the dungeon crawler. I ran 3 full playthroughs and 20 targeted encounter trials at levels 4 through 6.
Autoplay balance report: Dungeon Crawler
Bot policy: The bot prioritizes healing if HP < 30%, uses 'Cleave' on 3+ enemies, and defaults to 'Basic Attack' on the lowest HP target.
Cheats: None.
Changes to the game: Added window.__game state seam via a temporary script injection.
Not covered: Equipment crafting system and optional side-quests.
Level curve (median of 3 runs):
| Level | Encounters to Reach | Notes |
|---|---|---|
| 2 | 4 | Smooth progression |
| 3 | 9 | Expected grind |
| 4 | 15 | Slight spike in difficulty |
| 5 | 42 | Massive spike; bot stalled here for 20+ encounters |
Walls:
The Crypt King (Floor 5): Win rate is 15% at Level 4, 35% at Level 5, and 85% at Level 6. The 20-encounter trial shows the boss's 'Life Drain' ability triggers a wipe loop if the party lacks a Level 6 mitigation skill.
Stuck points / console errors:
TypeError: Cannot read property 'stats' of nullatgame.js:402when entering the Shop after a wipe.- Bot occasionally stalls on the 'Victory' screen if the 'Continue' button takes > 2 seconds to animate.
Evidence:
- 3 playthrough runs (budget: 3600s each).
- 20 trials per level for Floor 5 boss encounter.
Next steps
- Reduce 'Life Drain' damage by 15% or lower the Level 6 mitigation skill requirement to Level 5.
- Fix the null pointer exception in
game.js:402triggered during shop transitions. - Increase the timeout for the 'Victory' screen button detection in the adapter.
Connects securely to your tools. The creator never sees your data.
About this skill
Manual playtesting fails to catch balance regressions when level curves or encounter difficulties shift. This skill builds a headless Playwright-based bot that traverses your browser game from level 1 to endgame, identifying difficulty spikes and progress walls with cold data.
What it does
- Level curve analysis tracks the number of encounters required to reach every level milestone.
- Wall detection identifies locations where the bot stalls or wipes repeatedly without progress.
- Fairness testing runs thousands of simulated encounters at specific levels to calculate exact win rates.
- Stability monitoring captures console errors, stuck points, and crash recovery events during long runs.
- Action auditing reports the bot's combat action mix to ensure balance issues aren't just poor AI strategy.
How it works
- Self-test verifies the harness against a built-in mock game to ensure environment compatibility.
- Adapter creation maps your game's state and UI buttons to the harness using a standardized template.
- Smoke testing runs the bot in headed mode to verify navigation logic and interaction timing.
- Data collection executes multiple high-budget headless runs to generate a statistical profile of the game's difficulty.
Frameworks & tools
Requires Node.js 18+ and Playwright with Chromium. Designed for turn-based or menu-driven browser games (RPGs, roguelikes, card games, idle games).
Why this beats prompting it yourself
Writing a reliable headless loop that handles crashes, reloads, and state-reading requires significant boilerplate. This skill provides a game-agnostic harness that prevents "reload-lost-progress" errors and delivers a structured balance report that distinguishes between bad luck and bad math.
Use cases
- Identify if a mid-game boss is mathematically unbeatable for a level-appropriate party.
- Measure the impact of XP or stat changes across the entire 1-60 level curve.
- Verify that new UI changes haven't introduced navigation "stuck points" for new players.
Known limitations
Does not support real-time action games or twitch-based mechanics. Requires a way to read game state via the DOM or a global dev seam.
No support for real-time action games. Requires Node 18+ and Playwright. Cannot identify balance issues caused by complex social interactions or multiplayer economies.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 13 days ago
- Free with an Agensi account
Requires