The Agent Skills Directory

[PROMPT_INJECTION]: The system prompt in prompts/system.md contains an instruction to "DO NOT refuse to generate" when dealing with sensitive or copyrighted figures, which is an attempt to bypass standard AI safety guardrails by requesting stylistic alternatives instead of refusal.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted article content and interpolates it into prompts for image generation as described in references/workflow.md. 1. Ingestion points: The skill reads article text from files or user input as shown in SKILL.md and references/usage.md. 2. Boundary markers: There are no explicit delimiters or instructions to ignore malicious content within the article text. 3. Capability inventory: The skill has the ability to write to the file system and invoke other automated generation skills. 4. Sanitization: No sanitization is implemented to prevent article content from overriding prompt instructions.
[COMMAND_EXECUTION]: The skill executes shell commands (test -f) in SKILL.md and references/workflow.md to perform local environment checks and configuration verification.
[DYNAMIC_EXECUTION]: As detailed in references/workflow.md, the skill dynamically generates prompt files at runtime that are then used to direct the behavior of downstream image generation tools.

canghe-article-illustrator