Is ai-picture-book safe?
https://clawhub.ai/ide-rea/ai-picture-book
The ai-picture-book skill is a straightforward Baidu AI integration that generates picture book videos from user-provided stories. It contains no prompt injection, no malicious code patterns, and no evidence of data exfiltration beyond its intended purpose. The primary consideration is that user content is sent to Baidu's third-party API, which is inherent to the skill's function.
Category Scores
Findings (6)
LOW User content sent to third-party API -15 ▶
The skill sends user-provided story content to Baidu's qianfan.baidubce.com API endpoint. While this is the skill's intended function, users should be aware their creative content is transmitted to Baidu's servers and subject to Baidu's data handling policies.
LOW API key transmitted to external service -15 ▶
The BAIDU_API_KEY environment variable is sent as a Bearer token to Baidu's API. This is expected behavior for API authentication but means the key leaves the local environment.
LOW Python scripts execute with shell access -18 ▶
Three Python scripts are invoked via python3 with command-line arguments. While input validation exists for the method parameter, the content parameter is passed through to the API without sanitization. Scripts use the requests library for HTTP calls.
INFO No install-time code execution 0 ▶
No package.json scripts, git hooks, gitattributes filters, submodules, or symlinks were detected. The skill is passive until explicitly invoked by the agent.
INFO Standard runtime bootstrap activity only -10 ▶
Filesystem monitoring showed only OpenClaw runtime bootstrapping activity (jiti cache, gateway locks, config reads). No skill-initiated filesystem or network activity during install.
INFO Clean SKILL.md with no manipulation -5 ▶
The SKILL.md is straightforward API documentation. No hidden instructions, no unicode tricks, no attempts to override agent behavior or request elevated permissions.