对 AI 自动化流程做轻量红队演练，聚焦误用路径、边界失败和数据泄露风险。；use for red-team, ai, workflow workflows；do not use for 输出可直接滥用的攻击脚本, 帮助破坏系统.

This skill appears coherent and implements a local, resources-driven audit/report workflow using the bundled Python script. Before running: (1) avoid pointing the script or the agent at system-wide or sensitive directories (e.g., /, /etc, user home) unless you intend to scan them, because the script will read file contents you supply as input; (2) prefer running the script in an isolated workspace or container and use --dry-run when experimenting; (3) inspect outputs locally before sharing externally — while the script masks long secret-like tokens in reports, it still reads files that could contain secrets; (4) no credentials or network installs are required, but if you modify the skill to add network behavior, re-evaluate carefully. If you want stricter guarantees, review scripts/run.py directly or run it in a sandbox.

Type: OpenClaw Skill Name: ai-workflow-red-team-lite Version: 1.0.1 The skill bundle is a legitimate security auditing tool designed to perform light red-teaming and risk assessment on AI workflows. The primary logic in `scripts/run.py` performs static analysis, such as regex-based pattern matching for secrets and dangerous shell commands, and generates structured reports based on user-provided input or directories. The instructions in `SKILL.md` and `README.md` are strictly defensive, explicitly prohibiting the generation of harmful scripts or system destruction, and the code contains no network activity, obfuscation, or unauthorized data exfiltration mechanisms.