TokRepo — AI Asset Registry

Think of this as a package manager specifically for AI assets: prompts, skills, MCP configs, and scripts that agents can discover and install during tasks. It exposes 20+ tools covering the full lifecycle, from tokrepo_discover (turn a capability gap into ranked candidates with verification evidence) through tokrepo_install_plan and tokrepo_codex_install (safe, policy-checked installs with rollback) to tokrepo_harvest (package local work for reuse after a task). The trust model is thorough: content hashes, SBOM-lite, signature evidence, policy gates, and trust scores before anything touches disk. Useful when you want agents to pull vetted building blocks instead of reinventing boilerplate, or when you want to capture and share reusable artifacts from a session. Pairs with tokrepo init-agent to embed discovery hints in AGENTS.md and tool configs so future agents know to call it during planning.