Explore skills

Self-reflection, self-criticism, self-learning, and self-organizing memory. The agent evaluates its own work, catches mistakes, and improves permanently.

Senior Staff Engineer protocol that forces deep reasoning, prevents over-engineering, ensures code integrity with diff-aware changes, and validates library usage via web search. Never hallucinates APIs.

An opinionated guide for building dApps on Arbitrum using Stylus (Rust) and/or Solidity, covering local devnode setup, contract development, testing, deployment, and React frontend integration with viem. Use it when starting a new Arbitrum project, writing contracts, deploying to Arbitrum, or building a frontend.

This comprehensive security skill for AI-assisted projects operates in two modes: AUDIT, running a 50-point checklist with quantitative scoring, and PREVENTIVE, guiding secure code generation by consulting vulnerability knowledge bases. It covers OWASP Top 10, cloud-native security (Supabase RLS, Firebase Rules), and payment security.

This skill fetches and displays the latest trending content, generating Top items based on an estimated click-through rate rather than raw popularity. It supports querying current hotspots, service status, and grouped content, and can prompt users to set up scheduled tasks.

Use to search or retrieve information from Alibaba Cloud Bailian knowledge base, especially for knowledge base queries, document retrieval, semantic search, or any Bailian-powered information lookup.

Build stateful AI agents and agentic workflows with LangGraph in Python. Covers tool-using agents with LLM-tool loops, branching workflows, conversation memory, human-in-the-loop oversight, and production monitoring.

Performs deep, iterative multi-source research using NotebookLM and web search, enhancing results with domain boost and source credibility. It generates post-research artifacts (podcast, video, slides, quiz) and serves as a free alternative to Perplexity/Gemini Pro.

Mandatory security layer for any Claude Code session or action processing external content like WebFetch, WebSearch, GitHub reads, foreign repo READMEs, fetched PDFs, articles, doc sites, and MCP tool outputs from external services. It protects against prompt injection (output quarantine, injection pattern scan, tool-call injection defense, recursive fetch ban, hidden-Unicode tricks) and PII leaks in URLs.

Detects likely cognitive biases in product thinking, strategy documents, prompts, research summaries, meeting notes, user statements, and LLM outputs. Use for bias audits, classifying reasoning failures, or identifying the strongest bias candidates with evidence, distinctions, and countermeasures.

This skill is used to sanitize or prepare native crash stacks, typically from Chromium, for attachment to public issue trackers or code reviews. It transforms raw symbolicated stacks.

Use this skill to verify deployments after code changes to globally linked packages, running services, daemons, or scheduled tasks, especially when changes affect a globally installed CLI or a running service. Avoid using it for general development, CI setup, or initial deployments.