The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes external command-line tools to handle image generation, including bun, baoyu-image-gen, and the codex CLI. These commands utilize arguments—such as file paths for output and aspect ratios—that are derived from user input or derived slugs, which can be a vector for command injection if not properly sanitized by the runtime.
[REMOTE_CODE_EXECUTION]: The skill provides logic for discovering and executing a local script (main.ts) at runtime using a discovered path or environment variable ($BAOYU_CODEX_IMAGEGEN_BIN). Additionally, it suggests using npx -y bun as a fallback, which can download and execute code from the public NPM registry.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user content from source.md and interpolates it into structured prompts for an image generation backend. While it instructs the agent to strip credentials, it lacks robust boundary markers to prevent malicious instructions within the source content from hijacking the tool's behavior.
[EXTERNAL_DOWNLOADS]: The skill workflow involves potential downloads from well-known services (such as the NPM registry via npx). While these sources are generally considered safe under standard rules, the execution of downloaded content at runtime contributes to the skill's overall risk profile.

baoyu-infographic