RePrompter

Transform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "stru...

This skill appears to do what it says (improve prompts and optionally orchestrate multi-agent runs), but take these precautions before installing or invoking team-mode: 1) Repromptception mode expects tmux and an environment flag (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1) although they are not declared — verify your runtime provides tmux and that you understand enabling the experimental env var. 2) The skill's interview/auto-detection reads local project files (package.json, tsconfig, etc.) and writes temp files to /tmp — avoid running it from directories with secrets or production credentials, or use the documented opt-out. 3) Multi-agent execution can spawn tmux sessions and run multiple agents in parallel; confirm you trust the agent runtime and have appropriate isolation. 4) If you require guarantees, ask the author or maintainer to update the skill metadata to declare required binaries/env vars and to document precise filesystem scopes and permissions.

Type: OpenClaw Skill Name: reprompter Version: 7.0.0 The skill is classified as suspicious due to its extensive use of high-risk capabilities, specifically direct shell command execution via `tmux` and OpenClaw's `sessions_spawn` in `SKILL.md`. While these are intended for legitimate agent orchestration (e.g., `tmux new-session`, `tmux send-keys`, `ls`, `tmux kill-session`), the reliance on placeholders like `{session}`, `{taskname}`, `{domain}`, and `/path/to/workdir` creates a significant shell injection vulnerability if user input is not rigorously sanitized by the OpenClaw platform or the skill's internal logic. There is no evidence of intentional malicious exploitation (e.g., data exfiltration, persistence), but the inherent risk associated with these capabilities prevents a 'benign' classification.