diff --git a/README.md b/README.md
index c77a5a6..bca9d49 100644
--- a/README.md
+++ b/README.md
@@ -170,6 +170,7 @@ A curated list of Agent skill, awesome resources, tools, and tutorials for OpenA
 - [SwarmClaw](https://github.com/swarmclawai/swarmclaw) - Self-hosted multi-agent runtime that delegates to Codex CLI alongside Claude Code, Gemini CLI, OpenCode, Copilot CLI, Cursor Agent, Goose, Qwen Code, and Droid. Org chart view, schedules, runtime skills, persistent memory, and reviewed conversation-to-skill learning. MCP-native (server and client). Electron desktop app, CLI, and Docker.
 - [Alfred](https://github.com/luminik-io/alfred-os) - Self-hosted runtime that turns scoped GitHub issues into reviewed pull requests through autonomous Codex CLI and Claude Code agents. Per-firing git worktrees, label-driven state machine (agent:implement → agent:in-flight → agent:pr-open → agent:done), role-based engine routing across Codex and Claude, and Slack reporting. Python, MIT, macOS/Linux.
 - [Codex First Task Prompt Generator](https://ronnie2025.github.io/ai-agent-workbench-starter-pack/codex-first-task-prompt-generator.html) - Free web tool that turns a Codex CLI project goal into a scoped first-task prompt with constraints and acceptance checks.
+- [PickySteve](https://github.com/KernelLord/pickysteve) - Local-first skill router and context picker for coding agents: hybrid BM25 + embedding retrieval with cross-encoder rerank and an LLM judge selects the right skill, then assembles a minimal nonce-boundaried context bundle. Built-in fail-closed prompt-injection gate scans both the request and every retrieved document. Runs on local Ollama by default; MCP stdio server and OpenAI-compatible proxy wire into 18 coding agents including Codex CLI.
 
 ### Stat
 - [ccusage](https://github.com/ryoppippi/ccusage) - A CLI tool for analyzing Claude Code/Codex CLI usage from local JSONL files.