The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): In scripts/md/utils/languages.ts, the skill uses dynamic import() to load and execute JavaScript modules from a remote CDN (cdn-doocs.oss-cn-shenzhen.aliyuncs.com). The module path is dynamically constructed using the language identifier from Markdown code blocks, which could allow an attacker to execute arbitrary scripts in the agent's runtime environment.- [DATA_EXFILTRATION] (HIGH): The skill's instructions in SKILL.md specify reading files from the user's home directory ($HOME/.baoyu-skills/baoyu-markdown-to-html/EXTEND.md). This constitutes sensitive file access outside of the intended project scope.- [EXTERNAL_DOWNLOADS] (MEDIUM): The skill performs network requests to plantuml.com in scripts/md/extensions/plantuml.ts to fetch SVG content and dynamically imports the @antv/infographic package in scripts/md/extensions/infographic.ts.- [COMMAND_EXECUTION] (LOW): The skill relies on executing shell commands via npx and bun to run its main conversion logic. While this is part of its primary function, it involves spawning subprocesses for script execution.- [DATA_EXFILTRATION] (LOW): As an indirect prompt injection surface, the skill ingests untrusted Markdown data and produces HTML. Extension logic in infographic.ts and plantuml.ts injects content into the resulting HTML with minimal sanitization, which could lead to XSS or SSRF if the output is rendered in a browser or processed by subsequent tools.

baoyu-markdown-to-html