The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on extensive shell interaction to manage the webtoon creation process. It uses the codex CLI and standard Unix utilities such as ls, find, md5, awk, sort, and uniq to validate generated image files.
[COMMAND_EXECUTION]: Shell commands are dynamically assembled by interpolating data from external workspace files (e.g., ep{NN}_prompts.md). This creates a risk of command injection. The skill documentation explicitly acknowledges this by warning about the need for careful shell quote escaping when handling long prompts.
[COMMAND_EXECUTION]: The skill executes a specific shell script located at ~/.claude/skills/codex-image/scripts/codex_imagegen_batch.sh. This script is used to manage the batching and parallel execution of image generation tasks.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It ingests data from untrusted sources such as _workspace/04_visual/ep{NN}_prompts.md and style-bible.md. These inputs are used to build executable command strings for the codex tool. There are no robust boundary markers or sanitization mechanisms provided to isolate these external inputs from the command logic, increasing the risk that malicious instructions in the prompt files could influence the execution environment.

webtoon-panel-render