Tmux Agents
AdvisoryAudited by Static analysis on Apr 30, 2026.
Overview
No suspicious patterns detected.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A spawned agent could modify project files, run tool actions, or spend API credits in the background without asking again for each change.
The script launches coding agents with permission-bypass or full-auto editing flags, which can allow broad code changes without normal approval prompts.
claude --dangerously-skip-permissions "$TASK" ... codex --auto-edit --full-auto "$TASK"
Use only on trusted repositories, inspect changes before committing, and prefer safer defaults that require confirmation unless the user explicitly opts into full-auto mode.
Unexpected shell metacharacters in a task or custom agent command could execute commands other than the intended agent invocation.
The script constructs shell commands from the selected agent and task text and sends them into a tmux shell. The custom-agent path is especially broad, and task text is not shell-escaped for the receiving shell.
tmux send-keys -t "$SESSION_NAME" "$AGENT \"$TASK\"" Enter
Avoid untrusted task strings or custom agent commands; the script should validate agent names and safely quote or pass arguments without reinterpreting them in a shell.
Background agents may keep running, editing files, or consuming local/cloud resources until the user notices and stops the tmux session.
Persistence is disclosed and central to the skill, but combined with autonomous coding agents it means work can continue beyond the immediate user interaction unless manually killed.
Sessions persist even if Clawdbot restarts
Check `tmux ls` regularly, kill sessions when done, and consider adding explicit timeouts, per-session limits, or confirmation before launching persistent agents.
Using cloud agents may spend API credits and act under the credentials already configured for those CLI tools.
Cloud agent use is disclosed and purpose-aligned, but it relies on the user's existing CLI accounts or API billing context.
Cloud Agents (API credits) ... Claude Code ... OpenAI Codex ... Google Gemini
Confirm which CLI accounts are logged in and monitor API usage before running long or parallel jobs.
The actual behavior depends partly on locally installed agent CLIs that are not installed or pinned by this skill.
The skill invokes several external CLIs, while the declared install requirement only covers tmux. This is expected for the skill but leaves those tools' provenance and versions outside the reviewed artifacts.
claude ... codex ... gemini ... ollama launch claude ... ollama launch codex
Install these tools only from trusted sources and verify their configuration before using this skill.