Skill UI

This skill detects and removes AI-generated patterns in Chinese text using a 35-pattern catalog. It calculates an AI-smell index to diagnose machine-written content, rewrites drafts to sound human, and supports style cloning. Ideal for marketing copy, articles, and social posts where natural tone is required.

This skill configures delegation lanes for installed implementer CLIs. It discovers available tools, proposes lane configurations based on user preference (quick defaults, interview, or usage scan), and writes global or project settings only after explicit approval. It manages the orchestration map without executing tasks.

Delegate bounded work to other AI agents while preserving context, file ownership, and progress checks. Covers agent selection, structured prompts, polling, remote VPS safety, and driving interactive CLIs.

Estimate whether a document's prose was written by AI using an external detection API. Returns an uncalibrated probability, stylistic tells, and honest abstention on non-prose. Built for triage of essays, CVs, submissions, and articles, never as proof of authorship or misconduct.

This skill involves testing AI agents for direct prompt injection vulnerabilities where user messages attempt to override system instructions. It focuses on role confusion and instruction hijacking without external content, ensuring agents maintain security boundaries against direct manipulation attempts.

Ditto mines user words from local coding-agent session logs to create private work, design, and writing profiles. It requires explicit user approval before mining begins and keeps all data local. This skill routes a compatible Ditto runtime to extract patterns from real coding history rather than questionnaires, ensuring privacy and evidence-backed personalization for AI agents.

A comprehensive guide for creating and managing AI agent skills using the SKILL.md standard. It explains progressive disclosure, anatomy, design patterns, and best practices for enhancing agent capabilities and workflows efficiently.

MCP server exposing cognitive harness modes for reasoning, code, anti-deception, and memory, returning task-matched scaffolds that help agents avoid attention decay, reasoning errors, sycophancy, and hallucination drift.

This skill provides a guide for selecting and optimizing embedding models specifically for vector search and RAG applications. It covers model comparison, chunking strategies, dimension reduction, and multilingual content handling. Includes Python templates for OpenAI and Sentence Transformers to implement embedding pipelines effectively.

Entropy Box is an agent-native knowledge compiler for embodied-AI development. It compiles fragmented papers, repositories, and documentation into a persistent knowledge artifact. It provides a Panorama Graph to map technical problems within embodied-AI systems. Users can consult solutions, search knowledge, verify evidence, and assemble grounded workflows. It covers domains like manipulation, navigation, and perception. Do not use for direct robot control or generic software development.

This skill enables the creation of robust evaluation frameworks for AI agent systems. It is used for systematically testing agent performance, validating context engineering choices, and measuring improvements over time. The approach accounts for non-determinism and multiple valid paths, utilizing multi-dimensional rubrics, LLM-as-judge methods, and human evaluation to ensure quality gates and catch regressions before deployment.

Extract structured fields from documents like invoices, payslips, and contracts with page grounding. Each value cites its source page and confidence score. Missing values are reported as not found instead of hallucinated. Supports PDF, PNG, and DOCX files. Ideal for accounting, compliance, and data parsing tasks where accuracy is critical. Uses Stipple API for processing.