PySC2 (StarCraft II Learning Environment)
DeepMind's Python interface to the StarCraft II Learning Environment, exposing the game as an RL testbed for game-agent research.
PySC2 — StarCraft II Learning Environment
PySC2 is DeepMind's Python component of the StarCraft II Learning Environment (SC2LE). It wraps Blizzard's StarCraft II Machine Learning API as a reinforcement-learning environment, letting agents observe the game through simplified feature layers and act through the same interface a human player uses.
Key features
- Gym-style RL interface over the full StarCraft II game via the StarCraft II API
- Feature-layer observations (unit types, health, visibility, minimap) and a large structured action space
- Curated mini-games (build marines, defeat roaches, collect minerals) for focused skill learning
- Support for human-vs-agent, agent-vs-agent, and large-scale replay processing
- Rendering tools to visualize observations and inspect agent behavior
StarCraft II remains one of the hardest RL benchmarks because of imperfect information, huge action spaces, and long time horizons, making PySC2 a standard testbed for NPC and game-agent behavior research. Note: it requires an installed copy of the StarCraft II game plus the free map packs.
Curated mirror of the open-source PySC2 (Apache-2.0). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/pysc2-starcraft-ii-learning-environment/check). Click a policy:
Consume PySC2 (StarCraft II Learning Environment) programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/pysc2-starcraft-ii-learning-environment
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/pysc2-starcraft-ii-learning-environment/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add pysc2-starcraft-ii-learning-environment
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/pysc2-starcraft-ii-learning-environment/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "pysc2-starcraft-ii-learning-environment" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.