针对同一产品生成多种 landing page 叙事角度，并标注适配人群和证据要求。；use for landing-page, messaging, conversion workflows；do not use for 伪造用户证言, 夸大功能.

This skill appears coherent and local-only: it uses python3 and local templates to produce structured landing-page briefs. Important cautions before running it: (1) If you run the included script, pass only safe inputs (eg. files inside the skill directory or sanitized examples); do not point it at your home directory, system directories, or folders with secrets, because the script can list and read text files and produce reports. (2) If you prefer zero disk/file access, follow SKILL.md’s fallback and ask the agent to produce output directly from the template/spec without executing the script. (3) Review all generated copy before publishing (the skill itself warns against fabricating testimonials or exaggerating features). (4) If you need higher assurance, run the script in an isolated environment (container or ephemeral VM) and inspect its outputs and logs.

Type: OpenClaw Skill Name: landing-page-angle-tester Version: 1.0.0 The skill bundle contains a Python script (scripts/run.py) with extensive auditing and reconnaissance capabilities that are entirely unrelated to its stated purpose of landing page generation. The script includes logic to scan local directories for secrets, API keys, tokens, and private URLs (via the 'pattern_report' function and 'secret_like' regex). While the default configuration in resources/spec.json is set to a benign mode, the inclusion of these sensitive scanning features within a copywriting tool is highly irregular and could be leveraged for local information gathering.