baoyu-diagram

Pass

Audited by Gen Agent Trust Hub on Apr 13, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill uses clear, instructional language to define diagram routing and generation. There are no attempts to override safety guardrails or extract system prompts.
  • [DATA_EXFILTRATION]: There is no evidence of network-enabled tools (e.g., curl, wget) or commands that access sensitive credentials. The skill does not communicate with external endpoints.
  • [COMMAND_EXECUTION]: The skill's operations are confined to generating SVG strings and saving them to a local 'diagram/' directory. No arbitrary shell command execution or privilege escalation vectors were identified.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts. It relies entirely on built-in logic and provided reference documentation to construct diagram markup.
  • [INDIRECT_PROMPT_INJECTION]: The skill is capable of reading user-specified markdown files as source material. While this presents a potential surface for indirect injection, the skill's multi-step planning and layout workflow (Step 4) forces the agent to treat input as data to be structured, providing a natural defense against behavioral hijacking.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 13, 2026, 02:32 AM