ffsshhttiikk
ffsshhttiikk/opencode-agents-skills
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/ffsshhttiikk/opencode-agents-skills --skill reinforcement命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | reinforcement |
| description | Reinforcement learning fundamentals |
| license | MIT |
| compatibility | opencode |
| metadata | {"audience":"machine-learning-engineers","category":"artificial-intelligence"} |
Use me when:
Agent ──────▶ Action ──────▶ Environment
│ │
│◀─── State + Reward ◀──│
│ │
└─────── (Loop) ─────────┘
Goal: Maximize cumulative reward
import gym
import numpy as np
from collections import defaultdict
# Create environment
env = gym.make("CartPole-v1")
state = env.reset()
# Q-Learning Implementation
class QLearningAgent:
def __init__(self, n_actions, learning_rate=0.1,
epsilon=0.1, gamma=0.99):
self.q_table = defaultdict(lambda: np.zeros(n_actions))
self.lr = learning_rate
self.epsilon = epsilon
self.gamma = gamma
self.n_actions = n_actions
def choose_action(self, state):
if np.random.random() < self.epsilon:
return env.action_space.sample()
return np.argmax(self.q_table[state])
def learn(self, state, action, reward, next_state):
current_q = self.q_table[state][action]
max_next_q = np.max(self.q_table[next_state])
new_q = current_q + self.lr * (reward +
self.gamma * max_next_q - current_q)
self.q_table[state][action] = new_q
# Training loop
agent = QLearningAgent(env.action_space.n)
episodes = 1000
for episode in range(episodes):
state = env.reset()
done = False
while not done:
action = agent.choose_action(state)
next_state, reward, done, _ = env.step(action)
agent.learn(state, action, reward, next_state)
state = next_state