Local agent infrastructure in one stdlib-only Go binary

A zero dependency lightweight static analyzer + CVE scanner designed for adversarial-shape code in python to detect supply chain attacks before they reach your interpreter.

Zero-dependency, single-file Python implementations of popular libraries — benchmarked for performance parity | 零依赖单文件 Python 常用库实现，性能对标主流库

HUMMBL MCP Server

Plumego is a standard-library-first Go HTTP toolkit for explicit, maintainable, agent-friendly services, with zero external dependencies. It is designed to be simple, elegant, and efficient, making it ideal for small to large projects.

A plugin for Hermes Agent for meal planning and refrigerator management. It suggests recipes based on your ingredients and generates optimized shopping lists, combining conversational AI with a deterministic core without dependencies.

LLM-neutral skill that fetches a URL to a temp file so agents can pipe the path through rg / jq / awk instead of dumping the whole page into context. Same shape as pi-webfetch / opencode webfetch. Single-file Python 3 stdlib. Works on macOS, Linux, Windows.

TypeScript agent orchestration runtime for HUMMBL — kernel, router, adapters, control-plane, governance gateway. Multi-agent coordination with Base120 mental models.

Plan Mode orchestration tool — wraps Claude Code with structured Plan → Decompose → Execute workflow

Headless Chromium on ARM without Playwright, X11, or API keys. Termux, Raspberry Pi, embedded Linux. Public domain.

Seed-polymorphic reconnaissance engine. Six seed types in (IP, CIDR, Domain, ASN, CertFP, BannerString), a typed provenance graph out with rule-based exposure classification and drift detection.

Runtime safety primitives for AI agent systems — kill switch, circuit breaker, delegation tokens, governance bus. Zero dependencies. 58 tests. 7/10 OWASP ASI coverage.

Cross-platform CLI for the Linode API plus a Claude skill that drives it under explicit AI-safety constraints. Token never enters AI context.

HUMMBL Bibliography — provenance corpus, BibTeX citations, and position papers for Base120, HCC, BKI, and AI governance research.

Check what scopes your GitHub PAT actually has + what API calls it can/can't make. Stop guessing why 200 OK quietly does nothing.

AI deployment policy drift tracker

Pre-dispatch underwriting witness for delegated agent task slices: is this slice covered by an attested trust envelope and sufficient liability headroom? stdlib-only, deterministic, hash-chained decision ledger.

Edge robotics control evaluation engine checking telemetry deviations and sandboxed detour plans

Operational monitor for multi-agent AI. Per-action observations, per-agent rolling trends, baseline-relative alarms, named recommendations.

TrustLadder: an offline, stdlib-only measurement engine for a preregistered study of whether forcing an AI coding agent through a real enforcement gate cuts its shipped-defect rate. Signs hash-chained run-records, grades them blind behind a calibration gate, and runs frozen statistics that refuse to run until validity gates pass.