skills.rest
Lingjie-wang
Lingjie-wangCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

autorl

Plans and executes evidence-backed AutoRL workflows using Codex or Claude Code as executor.

Updated Jul 9, 2026
Search Tags:#verification#workflow orchestration#reinforcement learning#task card#autorl#executor brief#environment integration
One-click install
npx skills add https://github.com/Lingjie-wang/autoRL --skill autorl-lingjie-wang
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: autorl
Source: https://github.com/Lingjie-wang/autoRL
Command: npx skills add https://github.com/Lingjie-wang/autoRL --skill autorl-lingjie-wang

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Converting a multi-agent AutoRL pipeline into skill-based workflows is error-prone: tasks arrive vague, context gets bloated, and custom executors duplicate what coding agents already do. This Skill structures AutoRL work into bounded, evidence-backed stages so an external executor (Codex or Claude Code) can implement RL environment integrations and runtime artifacts without recreating an executor inside the workflow. ## Core Features & Use Cases - Task classification and task cards: Classifies requests as plan-only, implement, migrate, or debug, then builds a task card capturing intent, known facts vs assumptions, runtime contracts, evidence requirements, execution boundaries, and validation criteria. - Stage contracts instead of agent personas: Routes work through dedicated sub-skills for RL task clarification, evidence retrieval, environment integration, framework implementation, and independent environment verification emitting verification_report.json. - Bounded context handoffs: Gives the executor only the task card, evidence references, artifact contracts, and acceptance tests while keeping audit context outside the prompt. - Use Case: When migrating an existing multi-agent AutoRL codebase, use this Skill to map agent nodes to skills, prepare executor briefs, and validate produced artifacts against the task card before reporting success. ## Quick Start Ask the agent to use the autorl skill to turn your RL environment adaptation request into a task card with an executor brief and validation checklist.

Frequently Asked Questions about autorl

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I convert a multi-agent AutoRL pipeline into skill-based workflows?▼

Classify the request as a workflow migration, then map existing agent nodes to stage-contract skills such as rl-task-clarifier, rl-evidence-retrieval, rl-env-integrator, and rl-env-verifier. Remove custom executor responsibilities and let Codex or Claude Code act as the executor.

What is a task card in an AutoRL workflow?▼

A task card is built before execution and records user intent, the exact RL environment objective, known facts versus assumptions, runtime contracts, evidence requirements, execution boundary (generate-only, dry-run, or runtime), validation criteria, and the environment reuse policy.

Can this Skill run RL training itself?▼

No. The core rule is to treat Codex or Claude Code as the executor and never recreate an executor inside the workflow. The Skill structures tasks, context, handoffs, and validation; the external executor performs implementation and bounded test runs.

How are RL environment integrations verified?▼

The rl-env-verifier sub-skill independently checks environment integrations against the adapter contract in references/env-adapter-contract.md and emits a verification_report.json. Verification tiers and deliverables define what counts as an integrated environment.

When should I use the rl-task-clarifier sub-skill?▼

Use it when the RL task is vague and the executor would need to guess the environment, metric, budget, or runtime permissions. It runs a repeated clarification and ambiguity gate before evidence retrieval and strategy decisions begin.

Related Skills

Explore More Education & Research Skills→
rensjaspersrensjaspers

research

Gather facts and context from codebase and context-hub MCP.

Community
Intermediate
xlinh2301xlinh2301

arbor-agent-tools

Emulates Arbor tree, eval, merge, and report tools with a stdlib Python state script.

Community
Advanced
volter-aivolter-ai

develop

Implements GitHub issue acceptance criteria, commits evidence, and pushes a branch for PR review.

Official
Advanced
dbczumardbczumar
3

detect-framework

Detect Python agent frameworks from import patterns and map them to executor types.

Community
Intermediate