baoyu-infographic

Pass

Audited by Gen Agent Trust Hub on Nov 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content and is explicitly instructed in the workflow (Step 1.2 and Step 2) to preserve this data verbatim when generating structured content and prompts. This content is then interpolated into prompt templates (references/base-prompt.md) used to drive image generation tools.
  • Ingestion points: User-supplied text or files are saved to source.md in Step 1.2.
  • Boundary markers: The prompt template in references/base-prompt.md lacks strong delimiters or instructions to ignore embedded commands within the {{CONTENT}} and {{TEXT_LABELS}} placeholders.
  • Capability inventory: The skill can write files (SKILL.md Steps 1, 2, 5, 6) and execute shell commands/scripts (references/codex-imagegen.md).
  • Sanitization: While SKILL.md contains a rule to strip credentials and secrets, it does not sanitize against behavioral instructions or malicious payloads within the verbatim content.
  • [COMMAND_EXECUTION]: The skill requires executing external shell commands and scripts to perform its primary function of image generation. It specifically references calling baoyu-image-gen or a bundled wrapper via bun (references/codex-imagegen.md). While these are intended for the skill's purpose, they involve spawning subprocesses with parameters influenced by user-controlled prompt files.
  • [DYNAMIC_EXECUTION]: The skill utilizes dynamic logic to resolve the location of image generation tools at runtime. As described in references/codex-imagegen.md, it may search the plugin root or honor environment variables like $BAOYU_CODEX_IMAGEGEN_BIN to locate and execute TypeScript or binary wrappers.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 2, 2026, 06:28 AM