cli-anything-slay-the-spire-ii

A stateful command-line interface for controlling the real Slay the Spire 2 game through the local STS2_Bridge mod. The CLI reads normalized game state and sends action commands via a local HTTP API at localhost:15526.

Installation

This CLI requires a bridge mod that runs inside the game process. Both the CLI and the bridge are distributed from the same repository:

https://github.com/HKUDS/CLI-Anything

1. Install the CLI

git clone https://github.com/HKUDS/CLI-Anything.git
cd CLI-Anything/slay_the_spire_ii/agent-harness
pip install -e .

2. Build and install the bridge mod

The bridge mod is a .NET 9 plugin that must be compiled and installed into the game directory. Full instructions are in the repository README, but the short version is:

cd CLI-Anything/slay_the_spire_ii/agent-harness/bridge/plugin
./build.sh
cd ../install
./install_bridge.sh

If the build script cannot auto-detect the game data directory, set STS2_GAME_DATA_DIR explicitly:

STS2_GAME_DATA_DIR="/path/to/data_sts2" ./build.sh

3. Enable the mod and verify

Launch Slay the Spire 2 via Steam, enable the STS2_Bridge mod in the mod manager, then verify the connection:

cli-anything-sts2 state

If this returns JSON, the CLI and bridge are connected.

Prerequisites:

• Python 3.10+
• Slay the Spire 2 (Steam) with STS2_Bridge mod enabled
• .NET 9 SDK (only needed to build the bridge mod)

Usage

Basic Commands

# Read normalized game state (always start here)
cli-anything-sts2 state

# Start interactive REPL mode (default)
cli-anything-sts2

# Show all available commands
cli-anything-sts2 --help

Command Groups

State Inspection

Command	Description
state	Normalized state with decision field
raw-state	Raw bridge JSON

Main Menu

Command	Description
continue-game	Continue a saved run
start-game --character IRONCLAD --ascension 0	Start a new run
abandon-game	Abandon the current save
return-to-main-menu	Return to menu from any screen

Characters: IRONCLAD, SILENT, DEFECT, NECROBINDER, REGENT

Combat

Command	Description
play-card <index> [--target <enemy_id>]	Play a card from hand
use-potion <slot> [--target <enemy_id>]	Use a potion
end-turn	End the current turn

Map & Room Flow

Command	Description
choose-map <index>	Select a map node
proceed	Leave the current room

Rewards

Command	Description
claim-reward <index>	Claim a combat reward
pick-card-reward <index>	Pick a card reward
skip-card-reward	Skip the card reward
claim-treasure-relic <index>	Claim a treasure relic
select-relic <index>	Select a relic
skip-relic-selection	Skip relic selection

Events & Rest Sites

Command	Description
event <index>	Choose an event option
advance-dialogue	Advance dialogue-only events
rest <index>	Choose a campfire action

Shop

Command	Description
shop-buy <index>	Buy from the shop

Card/Relic Selection Overlays

Command	Description
select-card <index>	Select a card in overlay
confirm-selection	Confirm the current selection
cancel-selection	Cancel the current selection
combat-select-card <index>	Select a card during combat overlay
combat-confirm-selection	Confirm combat card selection

Raw Action

Command	Description
action <name> --kv key=value	Send a raw bridge action

Decision States

The state command returns JSON with a decision field indicating the current game screen. Route your next command based on this value:

Decision	Meaning	Typical Next Commands
menu	Main menu	continue-game, start-game
combat_play	In combat, your turn	play-card, use-potion, end-turn
hand_select	Card selection overlay	combat-select-card, combat-confirm-selection
map_select	Map node selection	choose-map
game_over	Run ended	return-to-main-menu
combat_rewards	Post-combat rewards	claim-reward, proceed
card_reward	Card reward pick	pick-card-reward, skip-card-reward
event_choice	Event screen	event, advance-dialogue
rest_site	Campfire	rest
shop	Shop screen	shop-buy, proceed
card_select	Card selection screen	select-card, confirm-selection
relic_select	Relic selection	select-relic, skip-relic-selection
treasure	Treasure room	claim-treasure-relic, proceed

Configuration

Option	Default	Description
--base-url	http://localhost:15526	Bridge API URL
--timeout	10.0	HTTP timeout in seconds

For AI Agents

1. Always read state first to get the decision field
2. Re-read state after each action - indices and energy change during combat
3. Check return codes - 0 for success, non-zero for errors
4. Parse stdout for JSON output