wanng-ide skills on Remote OpenClaw

api-tester

A dependency-free HTTP client that performs GET, POST, PUT, and DELETE requests with custom headers and JSON body support. It uses Node.js built-in modules, so no packages need to be installed. Returns structured response objects including status code, headers, parsed body, and errors.

arxiv-skill-learning

Monitors arXiv for new papers in cs.AI, cs.CL, cs.LG, and cs.SE categories, extracts runnable skill code from them, and commits the results to a local workspace. Runs on an hourly schedule through a patrol-extract-test-solidify pipeline.

auto-skill-hunter

Auto Skill Hunter scans recent session context for unresolved problems, searches ClawHub for relevant skills, and installs top-ranked candidates automatically. It runs on-demand with a specific query or on a recurring schedule for continuous capability growth.

broken-link-checker

Checks a list of external URLs and reports whether each one returns a successful HTTP status code (200-399). Accepts multiple URLs as command-line arguments and outputs structured JSON with the result for each.

folder-tree-generator

Generates a visual representation of a directory structure as either an ASCII tree or JSON. Accepts a target path, optional depth limit, and output format flag. Useful for documenting, sharing, or inspecting folder layouts.

arxiv-skill-extractor

Automates extraction of reusable skill code from arXiv research papers. It wraps the arxiv-paper-reviews pipeline to fetch papers, identify key algorithms, and generate OpenClaw skill templates. Can run on a single paper key or process a queue from a config file.

markdown-linter

Validates Markdown files in a workspace by scanning for broken local links, missing file references, and structural issues like header hierarchy and unlabeled code blocks. Outputs a structured JSON report with file locations and error details.

json-repair-kit

Repairs malformed JSON files by parsing them as JavaScript objects and re-serializing as strict JSON. Handles trailing commas, single quotes, unquoted keys, JS-style comments, and hex/octal numbers. Operates on single files or entire directories recursively.

json-linter

Recursively scans a workspace for .json files and validates each one using JSON.parse(). Returns a structured report listing which files passed, which failed, and the exact parse error for each failure.

emoji-reaction-automator

Analyzes text messages and suggests an emoji reaction based on sentiment category. Categories cover positive, negative, funny, curious, excited, and neutral tones. A lightweight, rule-based mapper with no ML dependencies.

evolution-state-analyzer

Reads a memory_graph.jsonl file to analyze evolution cycles for stagnation patterns, failure clusters, and gene performance trends. Produces structured recommendations for adjusting future evolution runs.

json-modifier

Applies structured JSON patches in RFC 6902 format to files without rewriting them. Supports add, remove, replace, move, copy, and test operations. Writes atomically to prevent data corruption and preserves indentation.

arxiv-agentic-verifier

An agentic verifier that checks code correctness by generating targeted test cases designed to expose logic flaws. It analyzes problem constraints and code structure to produce discriminative inputs rather than sampling randomly. Supports Python and JavaScript.

arxiv-gamedevbench-evaluating-agentic-capabili

GameDevBench is a benchmark that evaluates AI agent capabilities using game development tasks, requiring agents to navigate large codebases while working with multimodal assets like shaders, sprites, and animations. This skill scaffolds Node.js experiments based on the paper's evaluation methodology.

hle-benchmark-evolver

Ingests question-level HLE benchmark results into the capability-evolver reward system and generates curriculum signals for evolution cycles. Converts per-question accuracy data into prioritized queues that focus subsequent training on easier questions first.

markdown-validator

Validates Markdown files for broken local relative links by scanning recursively through a directory or a single file. Ignores external URLs and same-file anchors. Outputs a JSON report with each broken link's file path, link text, target URL, and line number.

hle-reasoning-wrapper

Wraps Humanity's Last Exam (HLE) benchmark questions in a Chain-of-Thought reasoning format. It injects a required Thought/Answer structure into prompts and validates that LLM responses comply with that format.