Public APIs Skill Creator

公共API/免费API SKILL生成器：从 public-apis/public-apis 自动检索免费 API，按功能推荐并给出最小可用调用示例（curl/Python/JS），并可自动生成自定义名称的 API skill。用户提到“公共API”“免费API”“public APIs”“找接口/找API”“生...

This skill appears to do what it claims (search public-apis, produce examples, and scaffold skills), but pay attention to two practical risks before running it: 1) Network probing risk: the optional --try probe issues a curl GET to the selected URL and prints response bytes. If an API entry (or a maliciously crafted link) points to an internal endpoint (localhost, cloud instance metadata, or other private services), the probe can retrieve and print sensitive data. Avoid using --try unless you trust the target URLs or run in an isolated environment; consider adding hostname filtering. 2) File write risk: the scripts cache data under ~/.cache and by default write generated skills into /root/.openclaw/workspace/skills. Review or override the --out-dir to a safe location you control, and do not run as root. Recommendations: run the scripts in a sandboxed VM or container, disable --try or inspect links before probing, set OUT_DIR to a non-privileged path, review generated call.sh/python_example.sh before executing them (they will call remote URLs and may require tokens), and avoid enabling autonomous agent execution if you do not want the agent making outbound network calls on its own. If you want to be safer, request that the author add host allowlists/denylists and make the OUT_DIR default configurable to the current user's workspace instead of /root.

Type: OpenClaw Skill Name: public-apis-skill-creator Version: 1.0.0 The skill is classified as suspicious due to multiple critical shell injection vulnerabilities (RCE risks) and prompt injection risks. Specifically, `scripts/create_skill.sh` directly interpolates user-controlled input (like API URLs and descriptions) into generated shell scripts and markdown files without proper sanitization, allowing for arbitrary command execution and prompt injection in the newly created skill. Similarly, `scripts/gen_usage.sh` and `scripts/solve_task.sh` (especially with the `--try` flag) directly interpolate API URLs into `curl` commands, creating shell injection vulnerabilities that could lead to RCE if exploited by a malicious input or a compromised API source.