Research & Data

Text handling helper systems built to isolate structural cross references inside unstructured financial reporting templates.

Coordinated context discovery systems built to execute recursive deep information research cycles completely autonomously.

Advanced custom embedding search routers configured to fetch localized information slices safely from cloud storage nodes.

Deep layout document analysis components engineered to extract structurally accurate data tables from unstructured papers.

Automatic Differentiation via Text. Gives agents backpropagation skills to optimize prompts and text-based system prompts.

Components and templates to build enterprise-ready deployment setups featuring advanced tool-use and search agents.

Implementation recipes demonstrating multi-source data ingestion routing for building specialized context knowledge pools.

Multi-tenant persistent memory server configurations built to host long-term state data for high-scale agent networks.

Build a RAG pipeline using natural language. Give your agent the skill to query complex vector stores through simple chat.

Parse and analyze AI agent skills (SKILL.md) — deterministic folder-to-JSON analysis library

Drop-in agent skill that verifies PDFs (fraud, tampering, AI-generation) via the VerifyPDF v2 API. SKILL.md + bash/python helpers.

Search & download ebooks from Z-Library — a portable Agent Skill (SKILL.md + stdlib-only Python CLI) usable by any AI agent (Claude Code, Codex, Cursor, Gemini CLI…) or from the shell.

面向 OpenClaw 的 Serper Google 搜索 skill。

A local Notion Data API skill for OpenClaw with helpers, validation, and redacted output support.

Unified LLM-agent skill for code review covering both security (CWE) and privacy (GDPR) through one detector-validator methodology. Ships the SKILL.md, a 128-case Python benchmark across 30 canonical categories with a held-out true-negative subset, and the evaluation harness from the SecPriv paper.

A structured-reasoning skill for AI agents. 5-stage loop (Observation -> Hypothesis -> Evidence -> Conclusion -> Verification) with weakest-link confidence, 2-axis severity, and cognitive bias check. See SKILL.md for the methodology.

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

Codex plugin for Alembic: MCP and skill runtime connecting Codex to local-first project memory, Guard, bootstrap, rescan, and Dashboard workflows.

Bilingual Codex skill for SRE incident response, Kubernetes troubleshooting, Prometheus alert triage, and postmortem drafting

Personal skills library for Claude Code and Cursor — drop-in SKILL.md files that give your AI agent reusable, invocable capabilities.

Index Claude Code conversations into SQLite via hooks, with local LLM summaries and a search skill that lets Claude recall past sessions

Claude Code skill: convert PDF to Markdown with LaTeX formulas via multimodal vision

Traceable research-paper workspace workflow for Codex skills

A memory-first AI agent that remembers why decisions were made — not just the last message. Runs local (Ollama), cloud (Claude · OpenAI · Gemini), or decentralized TEE. Graph memory, self-learning skills, multi-model routing, sandboxed tools. MCP · ACP · A2A. One Rust binary.