baoyu-cover-image

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructions in references/base-prompt.md explicitly direct the agent to circumvent potential safety or copyright refusals from the image generation model: 'If content involves sensitive or copyrighted figures, create stylistically similar alternatives
  • DO NOT refuse to generate.'
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data to generate prompts, creating a vulnerability surface.
  • Ingestion points: User-provided article titles, content summaries, and full article text are analyzed and interpolated into the generation prompt (SKILL.md, references/workflow/prompt-template.md).
  • Boundary markers: The prompt template uses standard Markdown headers (e.g., '# Content Context') to separate fields but lacks robust delimiters or explicit 'ignore' instructions to prevent embedded malicious content from hijacking the generation flow.
  • Capability inventory: The skill has the ability to write files to the local system (prompts/, refs/, EXTEND.md) and execute shell commands to invoke image generation backends.
  • Sanitization: There is no evidence of sanitization or escaping of the user-provided text before it is inserted into the final prompt file.
  • [DYNAMIC_EXECUTION]: The skill performs dynamic path resolution for its execution environment. references/codex-imagegen.md instructs the agent to 'Search the plugin root: walk up from this skill's directory looking for packages/baoyu-codex-imagegen/src/main.ts' to find and execute a wrapper script using bun.
  • [COMMAND_EXECUTION]: The skill requires the execution of various shell commands and CLI tools (e.g., baoyu-image-gen, codex, bun, npx) to facilitate image generation and environment setup as described in the workflow and backend resolution rules in SKILL.md and references/codex-imagegen.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 10:07 AM