Best new AI-agent and productivity resources: day 33

Building with AI agents? Then yesterday was a good day for you. There are practical picks for organizing work and creating documents too. Below are the best starting points, grouped by what they help you do. A quick caveat: these describe the listed projects and aren't guarantees of performance or security, so check each project's documentation before using it for consequential work.

Keep tabs on your agent sessions

Lost track of which terminal had that one coding session? Wake brings your AI-agent sessions into a single searchable desktop view, so you can find work and resume it without digging through terminal history. If you often run several coding sessions at once, start here. https://github.com/iAmCorey/Wake

Agentsmith is a model-agnostic harness with work-type profiles for agents like Claude, Codex and Gemini. Who's it for? Developers standardizing repeatable agent workflows, more than someone who only needs the occasional chat. https://github.com/PromptPartner/agentsmith

Context Loader takes a local Git repository and turns it into bounded Markdown or JSON context. Use it when your coding assistant needs a focused snapshot of a codebase, not an unfiltered dump of everything. https://github.com/xuanheng-tech/context-loader

Put a checkpoint between agents and their actions

Prismor is a self-hosted control plane that lets you observe, approve or block agent tool calls. If your agent can take actions that are hard to undo, its human-approval path deserves a close look. https://github.com/PrismorSec/prismor

Working on autonomous agents as a team? Microsoft's Agent Governance Toolkit combines policy, identity and sandboxing concepts in one place. Engineering and security teams are the audience here, not casual users. https://github.com/microsoft/agent-governance-toolkit

Govern places a signed decision gate in front of a deployment or agent workflow step, while Writ applies workflow gates and relevant rules inside Claude Code. In short, Govern is for teams managing CI and deployment decisions, and Writ is specific to Claude Code workflows. https://github.com/decionis/govern

Writ lives here: https://github.com/infinri/Writ

Two more joined this group: nono for agent runtime isolation, and Agent Control Standard for runtime-control conventions that don't depend on any one framework. Neither gets you out of reviewing permissions and testing a workflow yourself. https://github.com/nolabs-ai/nono

And Agent Control Standard is here: https://github.com/GenAI-Security-Project/agent-control-standard

Skills you can reuse, and knowledge workflows

SkillNet is about creating and evaluating reusable agent skills. Skillware, on the other hand, packages modular skills in Python. If evaluating skills matters most, go with SkillNet. If you need a Python framework to organize them, look at Skillware. https://github.com/zjunlp/SkillNet

Find Skillware at https://github.com/ARPAHLS/skillware

Claude Skills Collection is a broad set of skills and agents for several coding assistants. Don't install the lot. Browse selectively, and review individual skills before you add them to a tool that can access your files. https://github.com/borghei/Claude-Skills

Vercel Labs' Knowledge Agent Template is a starting point for an agent grounded in a file-system knowledge base. It suits developers who want to shape a template around their own documentation, rather than adopt a finished chatbot. https://github.com/vercel-labs/knowledge-agent-template

Not a developer? Everyday productivity picks

Not everything here is for developers. Bento is a portable, offline-oriented office suite for documents, spreadsheets and presentations. If you want to work on files without depending on a cloud office account, it's the most direct pick in this batch. https://github.com/nyblnet/bento

Open Supermarkets mixes command-line, MCP and agent skills for grocery search, comparison and checkout across supported providers. First check which countries and supermarkets it covers, and always review a purchase before you confirm it. https://github.com/abracadabra50/open-supermarkets

A few more specialist additions

Some picks are more niche. Sentio offers developers an email server and structured mail workflows for agents. Oracle VecDB Python SDK is about vector search and retrieval on Oracle AI Database. And Sphere SDK is an exploration of identity and transactions between agents. https://github.com/truespar/sentio

Oracle VecDB Python SDK lives here: https://github.com/oracle/vecdb-python-sdk

And Sphere SDK is here: https://github.com/unicity-sphere/sphere-sdk

DomA explores browser automation. UniFace brings several face-analysis tasks into one Python interface. And ainovel-cli experiments with writing novels using multiple agents. Only reach for these when their particular workflow matches your project. https://github.com/dom-actions/doma

Find UniFace at https://github.com/yakhyo/uniface

And ainovel-cli is at https://github.com/voocel/ainovel-cli

Rounding things out are Prismor's neighboring research and discovery resources: Awesome Multi-Agent Orchestrators, Awesome TypeSafe Jev, Fable 5 Opus 5.5 Orchestrator and DSH NextTavern. Want to see how the new entries compare with older alternatives? Browse the entire hand-curated directory. https://bestproductivityresources.com/directory