Search

Intent-Oriented Programming: FastAPI middleware + formal multi-agent contracts with SEPA and Assume-Guarantee guarantees

Markdown formatting skill for AI coding agents — formats any content into clean, beautiful .md files compatible with Typora/GitHub/VS Code

Observability SDK for AI agent workflows — automatic Anthropic and OpenAI instrumentation

Install Kushi — multi-source project evidence agent with Comprehensive Structured Capture (CSC) into weekly-only files across Email, Teams, OneNote, Loop, SharePoint, Meetings, CRM, ADO. Meetings retain a sibling verbatim/ audit folder. WorkIQ-only for M3

The open-source accountability layer for AI agents. Trace every action. Explain every decision. Control what matters.

Dual-mode interactive CLI prompts — terminal UI for humans, JSON line protocol for AI agents

db9.ai service adapter for Sandbank — serverless PostgreSQL for AI agents

AST mutation CLI for AI agents. `mutates` is a thin client over `@mutates/core`; a per-project daemon owns a single `ts-morph` `Project`, mints opaque `@nN` node refs, applies structural edits in memory, and saves with mtime+size conflict detection.

Crash detection, auto-repair, and health monitoring for AI agent gateways. Watches your OpenClaw or Hermes gateway, detects crashes, runs instant local repairs, and reports to Agent Hospital for AI-powered diagnosis.

Downcity Agent 运行时 — 执行壳，通过 RPC/HTTP 调用 City 的能力

PARALLAX ENGINE plugin for OpenCode -- protocol enforcement, friction-loop verification, mode switching (plan/build/debug/horizon), trace recording, coherence scoring, CI gate, PR-ready trace artifacts, and autonomous long-horizon supervision

AI-native development framework — asset memory, SDD gate, dual-track verification, self-iteration hooks

AI agent audit platform SDK. Captures telemetry from LangChain, CrewAI, AutoGen, LlamaIndex, and Haystack agents.

Deterministic combinatorial coverage for agent workflows and constrained variant surfaces.

Rebuild and compare Claw Code parity with harness tools for testing, analysis, and archive review

Claude Patcher -- MCP tools that extend Claude Code beyond built-in capabilities. Fix encoding, convert formats, detect duplicates, batch rename, archive, and more.

United Shared Memory Client -- cross-agent persistent memory access layer for LLM systems

LLM swarm intelligence toolkit — 5 parallel execution patterns for orchestrating multiple LLM instances

Testing framework for LLM operating systems - B/O/E test methodology

Self-hosted AI research & knowledge stack: Ollama + n8n + Rinnsal + KnowledgeDigest