## Core Insight agentskills.io is an open standard for agent skills that works across Hermes, Claude Code, Cursor, Codex, and other agent platforms. Skills written to this spec are portable — you write once and run on any compatible agent. This is the "marry the harness" layer. When skills are standardized, switching the underlying model or agent platform doesn't require rewriting your capabilities. The skill is the durable asset; the model is the replaceable substrate. ## Ecosystem Evidence (Mar 2026) - **wondelai/skills**: 250+ stars, cross-platform library for Claude Code and agentskills.io platforms - **Anthropic-Cybersecurity-Skills**: 734 security playbooks mapped to MITRE ATT&CK (3.6K stars) - **Chainlink agent skills**: Official oracle network skills on the spec - **Black Forest Labs/skills**: FLUX image generation skills - **skilldock.io**: Cross-platform skill registry (Hermes, Claude Code, OpenClaw) - **litprog-skill**: Works across Claude Code, OpenCode, and Hermes simultaneously ## Bridges Already Built - **evey-bridge-plugin**: Claude Code ↔ Hermes context sharing and task handoff - **hermes-agent-acp-skill**: Multi-agent delegation across Hermes, Codex, and Claude Code — routes subtasks to best-suited agent - **zouroboros-swarm-executors**: Local executor bridge for Claude Code + Hermes integration ## Cross-Domain Applications - **Personal infrastructure**: Your 199 Claude Code skills could be ported to agentskills.io format, making them model-portable — directly addresses the model lock-in frustration - **Product opportunity**: A "Claude Code skills → agentskills.io converter" tool would serve the exact audience feeling the same lock-in pain (S021 demand signal) - **Career positioning**: agentskills.io expertise is rare and immediately valuable — the standard is new, the ecosystem is growing fast ## Source - [[Awesome Hermes Agent Ecosystem]] — github.com/0xNyk/awesome-hermes-agent (Mar 30, 2026)