article-writing
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites and search results via
scripts/collect_serp_outlines.py. It mitigates risks by instructing the agent to ignore instructions in external content, using clear trust boundary markers in generated reports, and applying markdown escaping to prevent content from breaking the agent's context. The skill also uses independent subagents for section drafting, which provides isolation for processed data. - [EXTERNAL_DOWNLOADS]: The skill fetches data from the DataForSEO API and various public URLs. The implementation in
scripts/collect_serp_outlines.pyincludes a robust SSRF (Server-Side Request Forgery) protection layer that validates target IP addresses are public and global before connecting, and it includes logic to strip sensitive headers like Authorization during cross-origin redirects. - [COMMAND_EXECUTION]: The workflow involves running a local Python script with user-provided keywords. The instructions provide explicit sanitization rules, such as keyword slugification that removes shell metacharacters and directory traversal sequences (e.g., ../), to ensure safe execution.
- [CREDENTIALS_UNSAFE]: The skill requires API credentials for DataForSEO but follows best practices by requesting they be set as environment variables and explicitly warning the user never to paste secrets into the chat or print their values.
Audit Metadata