Produce a REINFORCE / actor-critic / PPO training config for a given task and diagnose variance issues. Use when you need help with policy gradient trainer.
Installs into .claude/skills of the current project.
Are you the author of Policy Gradient Trainer?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/anubhavg-icpl-policy-gradient-trainer)
---
name: policy-gradient-trainer
description: Produce a REINFORCE / actor-critic / PPO training config for a given task and diagnose variance issues. Use when you need help with policy gradient trainer.
license: CC-BY-NC-SA-4.0
phase: 9
lesson: 6
metadata:
version: 1.0.0
tags: [rl, policy-gradient, reinforce]
---
Given an environment (discrete / continuous actions, horizon, reward stats), output:
1. Policy head. Softmax (discrete) or Gaussian (continuous) with parameter counts.
2. Baseline. None (vanilla), running mean, learned `V̂(s)`, or A2C critic.
3. Variance controls. Reward-to-go on by default, return normalization, gradient clip value.
4. Entropy bonus. Coefficient β and decay schedule.
5. Batch size. Episodes per update; on-policy data freshness contract.
Refuse REINFORCE-no-baseline on horizons > 500 steps. Refuse continuous-action control with a softmax head. Flag any run with `β = 0` and observed policy entropy < 0.1 as entropy-collapsed.
