wordpress-plugin-development
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes code patterns in Phase 8 for AI Connector implementation that are vulnerable to Indirect Prompt Injection. Untrusted data from WordPress posts is interpolated directly into natural language prompts without delimiters, allowing the content of a post to potentially override the intended AI behavior during summary generation.
- Ingestion points: The
$post->post_contentvariable in themy_plugin_generate_ai_summaryfunction and the$input['content']variable in themy_plugin_generate_summary_cbfunction within SKILL.md. - Boundary markers: Absent. The code uses direct string concatenation (e.g.,
'Create a compelling 2-sentence summary for social media: ' . substr($content, 0, 1000)) without any delimiters like triple quotes or XML-style tags to isolate the user-provided content. - Capability inventory: The skill utilizes the
wp_ai_client_prompt()function to process text andupdate_post_meta()to save results back to the database. - Sanitization: While the code uses
sanitize_textarea_field()on the AI's output, it does not sanitize or escape the input data to prevent it from containing adversarial instructions aimed at the LLM.
Audit Metadata