Zellij Terminal Workspace

Remote-control zellij sessions for interactive CLIs by sending keystrokes and scraping pane output.

This skill appears to do what it says: control zellij sessions and scrape pane output. Before installing, ensure you have zellij and jq installed (SKILL.md expects both). Confirm where CLAWDBOT_ZELLIJ_DATA_DIR will point (default is ${TMPDIR:-/tmp}/moltbot-zellij-data) — the scripts operate on that directory and will list/delete sessions there. Be cautious when launching or automating commands inside panes (the examples run coding agents and arbitrary shell commands): those commands run with your account privileges and can modify files or exfiltrate data if untrusted. Note the minor metadata mismatch: the registry lists no requirements while SKILL.md expects zellij and jq — this is likely documentation drift, not a security problem, but you may want to verify requirements before use.

Type: OpenClaw Skill Name: zellij Version: 1.0.1 The OpenClaw AgentSkills skill bundle for 'zellij' is classified as benign. All files, including the SKILL.md and helper scripts, are clearly aligned with the stated purpose of managing and interacting with zellij terminal multiplexer sessions. The SKILL.md provides instructions for the agent to use zellij for legitimate tasks like running Python REPLs or orchestrating coding agents, including examples of sending commands within zellij panes. The helper scripts (cleanup-sessions.sh, find-panes.sh, find-sessions.sh, wait-for-text.sh) perform standard zellij session management operations. There is no evidence of intentional harmful behavior such as data exfiltration, malicious execution, persistence, or prompt injection with a malicious objective.