article-rewrite
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a highly structured environment for processing articles. It includes multiple validation steps (e.g.,
check.mjs,doctor.mjs) to ensure output quality and environment compatibility. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external articles (
article.md). While this is a vulnerability surface, the documentation inreferences/writing-guide.mdexplicitly warns the AI about potential injection markers like 'ignore previous instructions' in external sources. The use of a structural validation script (check.mjs) further mitigates risks of schema confusion. - [CREDENTIALS_UNSAFE]: The skill handles API keys for image generation (OpenAI, Gemini) through standard configuration files (
~/.config/article2media/config.json) rather than hardcoding them, which is a secure practice for CLI-based tools. - [DYNAMIC_EXECUTION]: The skill allows the use of hand-written SVG for complex charts as a last resort. The documentation includes explicit security constraints for this feature, prohibiting
<script>tags, event attributes, and external links, effectively neutralizing the cross-site scripting (XSS) or code injection vectors often associated with SVG processing. - [COMMAND_EXECUTION]: Command execution is strictly limited to an allowed set of local Node.js scripts via the
allowed-tools: Bash(node ${CLAUDE_SKILL_DIR}/scripts/*)directive, preventing the execution of arbitrary or malicious shell commands.
Audit Metadata