herdr provides a terminal-native agent multiplexer that exposes workspaces, tabs, and panes. This skill documents how an assistant can control herdr via its CLI to inspect panes, create tabs, split panes, run commands, wait for output or agent status, and coordinate multiple agents and processes in one terminal. Concrete recipes cover running servers, running tests in isolated panes, waiting for readiness, and spawning additional agents.
Use this skill when the assistant is running inside a herdr session (HERDR_ENV=1) and needs to: manage panes/tabs/workspaces, run commands in sibling panes, wait for specific output before continuing, check another agent's status, or orchestrate parallel tasks in the terminal. It's aimed at developer workflows (builds, tests, log watching, agent coordination).
Likely compatible with CLI-oriented agents and developer-focused clients that can run shell commands (Codex, Claude Code, Cursor, CLI-based assistants).
herdr is a terminal multiplexer for agents. The SKILL.md is thorough documentation for controlling herdr workspaces, tabs, panes, and agents via CLI. The 8 bundled scripts are project build/test/dev tools (changelog management, terminal key capture, vendoring), not agent-executable skill scripts. None ran successfully in audit — 3 require interactive TTY, 2 have module import issues, 2 need project-specific context, and 1 requires subcommand args. The SKILL.md itself is well-written and useful for herdr users.
The scripts are developer tooling for the herdr project itself, not general-purpose scripts that an agent would execute as part of the skill workflow. The actual skill is the SKILL.md documentation which teaches an agent how to control herdr. No security concerns — all scripts are transparent, no network calls to suspicious hosts, no credentials, no destructive commands without guards.