This skill guides an agent through the tower defense loop in any engine: choosing a map model (fixed lanes, maze or flow-field open field), writing towers, enemies and waves as data, and tuning the economy that links them. Its core question is whether the damage per second a player can afford keeps pace with the hit points each wave sends.
A bundled Python script plays the economy forward and reports the pressure of every wave, so walls and runs of trivial waves show up before a human playtest. The skill also lists targeting modes, sell refunds, early-call bonuses and a set of gotchas for stuck enemies, drifting timers and unfair specials.
What You Get
- A workflow from map model through wave data, curve tuning and session pacing
- An economy table with starting values for wave 1, bounties, leaks, sell refunds, interest, upgrades and targeting
- A curve checker script that prints each wave’s pressure and flags walls and trivial runs
- A wave-data pattern and a list of gotchas for pathing, timers, corners and unfair specials
Common Use Cases
- Balancing the waves and gold economy of a Bloons- or Kingdom Rush-like game
- Fixing late waves that are impossible, trivial or dominated by one tower
- Choosing targeting modes and adding enemy types that counter lazy builds
- Validating maze placements so towers never cut the spawn off from the goal
What Makes It Unique
Most game-design skills describe balance in prose. This one ships a small checker that plays the gold economy forward against the wave list and reports a pressure number per wave, so the agent can quote measured figures. It works in any engine and hands pathfinding, tile maps and data resources off to named sibling skills.
Under The Hood
The agent loads the SKILL.md and, when it needs depth, a balancing reference on path representation, targeting code and HP scaling. For the curve check it runs a local Python script, using only the standard library, on a JSON wave spec you write; the script makes no network calls and exits with a code the agent can read. The workflow also tells the agent to write and run small tests for travel time, maze validation and targeting in your project, which means running commands and creating test files.