Skip to content

AI/ML · Global library

Reward Hacking Preventer

Design robust reward functions and evaluation frameworks that prevent reward hacking and specification gaming

CodexClaude CodeKimi Codeorchestrator-mcp

Best use case

Use Reward Hacking Preventer when you need to design robust reward functions and evaluation frameworks that prevent reward hacking and specification gaming, especially when the work is driven by reward hacking and specification gaming.

Trigger signals

reward hackingspecification gamingreward shapingproxy gamingincentive misalignment

Validation hooks

gaming-detectionreward-balance

Install surface

Copy the exact command path you need.

Inspect

pip install "orchestrator-mcp[dashboard]"
orchestrator-mcp skills show reward-hacking-preventer

Use

orchestrator-mcp skills export reward-hacking-preventer --to ./skillforge-packs
# copy the exported pack into your preferred agent environment

Export

cp -R skills/reward-hacking-preventer ./your-agent-skills/reward-hacking-preventer
# or open skills/reward-hacking-preventer/SKILL.md in a markdown-first client

File patterns

*.pyrl*.pyreward*.py

Model preferences

claude-opus-4gpt-4oclaude-haiku-3

Related skills

Adjacent packs to compose next.

AI/MLGlobal library

Agent Lifecycle Manager

Open pack

Manage complete agent lifecycles from initialization through graceful shutdown with health monitoring, scaling, and resource optimization

CodexClaude Code
AI/MLGlobal library

Agent Memory Designer

Open pack

Design short-term, long-term, and episodic memory layers for agents without turning retrieval into an unbounded context leak.

CodexClaude Code