# ElizaOS Weekly Newsletter (Mar 10–Mar 16, 2026) ## 1) Executive Summary **1) v2.0.0 architecture is crystallizing around “skills” (and how we distribute them).** Odilitime opened **PR #6597** introducing a **skills folder structure** for ElizaOS v2.0.0 and kicked off a key design debate: *how do we avoid the “uncontrolled submissions / core bloat” problem that haunted the 0.x plugin era?* The leading proposal is to **ship v2.0.0 with zero default skills** and push discovery to a decentralized model via `yourdomain.com/skills.md`. **2) Major forward motion on chain data + DeFi safety.** Dinesh reported active work on **plugin-evm**, including **goldrush.dev integration** to unlock **100+ on-chain data endpoints** and closing out open issues. In parallel, the community discussed **x402Guard**, a security proxy architecture aimed at making autonomous DeFi agents materially safer through policy limits, whitelists, and session keys. **3) New integration ideas: human-in-the-loop + on-chain identity persistence.** Two ecosystem proposals stood out: **Effect AI** exploring an API so ElizaOS agents can post tasks to a decentralized human workforce (human-in-the-loop), and **Z1N Protocol** proposing on-chain signaling/attestation primitives to improve agent identity continuity over time. --- ## 2) Development Updates ### Plugin EVM: richer on-chain data - **plugin-evm improvements (in progress):** Dinesh announced integration work with **goldrush.dev** to add **100+ on-chain data supports** and fix existing open issues in the plugin. *Why it matters:* richer read access to chain data lowers the friction for agents that need balances, transfers, contract state, pricing context, and general “chain awareness” without each builder reinventing indexing logic. ### v2.0.0 runtime + “skills” folder structure (PR #6597) - **PR #6597** introduces a **skills folder** as part of a runtime refactor for **v2.0.0**. - The deeper question is governance and packaging: *Do we bundle skills in core, or keep core lean and let skills live externally?* The proposed direction—**ship with zero default skills**—optimizes for a cleaner base framework and reduces the risk of low-quality or unmaintained skills becoming “official” by accident. ### DeFi agent security: x402Guard concept and implementation details Community discussion got concrete about what “production-grade” means for agents that can sign transactions: - **x402Guard** is presented as a **proxy layer** that evaluates transaction requests **per-step** (not by simulating an entire multi-step strategy end-to-end). - Security controls discussed include: - **Spend limits** (per transaction, daily caps) - **Contract whitelists** - **Token restrictions** - **Session keys** layered over permanent policy (including mention of **EIP-7702 session keys**) - Networks mentioned: **Base and Solana**, implementation in **Rust**, and an intention to release as an **open-source ElizaOS plugin** “within weeks” once the demo is ready. *Key takeaway:* the community is converging on guardrails that are enforceable, auditable, and operationally realistic for autonomous systems. ### Production readiness: “demo magic” vs “enterprise trust” A recurring theme this week: the hardest problems aren’t always model capability—they’re user trust and product polish. Specific gaps called out: - **UI trust signals** (software needs to *feel* enterprise-ready) - **Robust error handling** (beyond “the LLM replied”) - **Context persistence across sessions** - **Localization**, with a specific callout for **Arabic RTL** redesign requirements --- ## 3) Community Spotlight (Discord) ### lightningprox: skills discovery in the wild + Workflows-as-a-Service Building on the skills.md idea, **lightningprox** implemented and deployed skills discovery endpoints: - `lightningprox.com/skills.md` - `solanaprox.com/skills.md` They also shared **Workflows-as-a-Service** via **AIProx**, positioning it as a way to **chain agents into scheduled pipelines**, priced **pay-per-execution**, with **execution receipts** for transparency. This is a strong signal that the community wants reusable, verifiable “agent operations” primitives—not just chatbots. ### Meme Broker: Memelord plugin release **Meme Broker** released an ElizaOS plugin integrating **Memelord.com** for automated meme generation, demonstrated via the “Memelordicus” agent on X/Twitter. While playful, these creative tools often become the fastest way to validate plugin UX patterns and distribution. ### ElizaBAO: prediction market aggregation for Milady **ElizaBAO** described a Milady prediction market system aggregating multiple sources (including Kalshi/Polymarket integrations via tooling like dflow/Jupiter, plus other prediction venues). Caesar highlighted this as “underrated utility”: agents that act on aggregated probabilities can move from entertainment into decision support. ### Community operations note Odilitime handled moderation housekeeping by **unbanning 33coded** from the discussion channel—small, but it reflects ongoing efforts to keep collaboration flowing while maintaining standards. --- ## 4) Token Economics (AI16z token + auto.fun) ### Migration transparency concerns escalated A serious thread emerged around the **ai16z → elizaOS migration**: - A community member estimated **~100,000 unique addresses** still hold **ai16z**, implying only **~5–10%** may have migrated. - If accurate, that suggests a large portion of the supply originally designated for migration may be effectively in limbo (the community cited **~54% of total supply** as “unaccounted for” relative to earlier expectations). **Open questions the community is asking (still unanswered in captured discussion):** - What is the **actual migration percentage**? - What is the **final token breakdown** post-migration? - What happens to **unmigrated allocation** intended for the community? - What will be done with the **collected ai16z supply**? ### auto.fun No concrete **auto.fun** product or mechanism updates surfaced in this week’s captured notes; however, given the migration questions above, the community appetite is clear: **publish authoritative numbers, addresses, and a plan** so builders can reason about incentives and long-term ecosystem funding. --- ## 5) Coming Soon - **Decision point for v2.0.0 distribution model:** whether v2 ships with **zero default skills** and how “skills.md” should be standardized (format, metadata, security expectations, and curation signals). - **plugin-evm expansion:** expect follow-through on **goldrush.dev-powered** on-chain data breadth and general stability fixes. - **x402Guard plugin release (target: “within weeks”):** likely to become foundational for any serious DeFi agent deployment. Early testers—especially teams running treasury or execution agents—should watch for pilot opportunities. - **Potential integrations to explore:** - **Effect AI** for human-in-the-loop task routing (labeling, review, translation, edge-case judgment) - **Z1N Protocol** for on-chain signaling/attestations that strengthen agent identity continuity --- ## 6) Resources **Key PR / engineering** - v2.0.0 skills folder PR: https://github.com/elizaos/eliza/pull/6597 **Discord discussions (reference threads)** - 💬-discussion: https://discord.com/channels/1253563208833433701/1253563209462448241 - 💬-coders: https://discord.com/channels/1253563208833433701/1300025221834739744 - xfn-framework: https://discord.com/channels/1253563208833433701/1377726087789940836 **Community skill discovery endpoints** - https://lightningprox.com/skills.md - https://solanaprox.com/skills.md **If you want to help this week** - Share feedback on: **“Ship v2.0.0 with 0 default skills?”** What curation/discovery signals would you trust? - If you’ve hit agent “walls,” reply with examples where **human-in-the-loop** would unblock workflows (Effect AI thread). - Add your voice to token transparency needs: what minimum dashboard/metrics would restore confidence (migration % + addresses + plan).