The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands ('test -f') in Step 1.1 of the workflow to verify the existence of the 'EXTEND.md' configuration file in both the project-local directory and the user's home directory.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content and processes it for use in an image generation prompt. It implements a verbatim data policy that preserves source text exactly, creating a potential vector for embedded instructions to reach the image generator. 1. Ingestion points: 'SKILL.md' (Step 1.2) via file path or direct paste. 2. Boundary markers: 'references/base-prompt.md' uses simple double-brace placeholders ({{CONTENT}}) without additional delimiters or instruction-ignore warnings. 3. Capability inventory: 'SKILL.md' uses Bash for file checks, performs multiple file-write operations, and invokes external image generation skills. 4. Sanitization: Absent; the skill explicitly requires that all data be copied word-for-word.

canghe-infographic