The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Safety filter bypass instruction. The file references/base-prompt.md contains a directive stating 'DO NOT refuse to generate' when content involves sensitive or copyrighted figures, explicitly instructing the AI to circumvent standard refusal protocols by creating stylistic alternatives instead.
[PROMPT_INJECTION] (LOW): Indirect Prompt Injection surface. The skill ingests untrusted article content (Step 1) and processes it to form prompts for downstream image generation tools (Step 5) without utilizing sanitization or explicit boundary markers, which could allow malicious instructions within a processed article to hijack the generator's output.
[COMMAND_EXECUTION] (LOW): Dynamic configuration through extension files. As described in SKILL.md, the skill supports loading an EXTEND.md file from the local project or home directory to override default behaviors and styles, creating a mechanism for logic modification if the local filesystem is compromised.

cover-image