publish-blog-article
Pass
Audited by Gen Agent Trust Hub on Aug 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill reads content and frontmatter from files in 'content/posts/' (e.g., 'publish.juejin.category', 'publish.x.text') to parameterize syndication tools. This creates an indirect prompt injection surface if the blog content is influenced by untrusted sources. 1. Ingestion points: Markdown files in the content directory. 2. Boundary markers: No delimiters or specific ignore instructions are used for the parsed data. 3. Capability inventory: Git push operations, pnpm scripts, Hugo builds, Wechatsync MCP calls, and browser control. 4. Sanitization: The skill does not specify validation or escaping for the metadata used in API calls.
- [COMMAND_EXECUTION]: The skill requires the ability to execute shell commands, including 'git push', 'hugo' builds, and repository-specific 'pnpm' scripts. These capabilities are directed by instructions parsed from local repository files without a secondary verification layer.
Audit Metadata