processon-ai-skill
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data (such as meeting notes, documents, and web content) to generate structured mind maps and flowcharts.
- Ingestion points: The skill ingests natural language, documents, and images for processing into diagrams, as documented in
references/mindmap.mdandreferences/flowchart.md. - Boundary markers: The agent is instructed to use shell-level heredoc delimiters (
POB64) when communicating with local scripts to prevent command injection, but there are no specific natural language delimiters or "ignore embedded instructions" warnings for the data being summarized. - Capability inventory: The skill can perform network operations to ProcessOn's official endpoints and can write/update files in a user's account (
createProcessOnFile,updateProcessOnFile). - Sanitization: Content is base64 encoded for shell safety during transport, but no semantic filtering or sanitization of the input text is applied prior to AI summarization.
Audit Metadata