The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted user content verbatim into a prompt template that is later processed by an image generation skill.\n
Ingestion points: User-provided content passed via file path or raw text input, which is saved to source.md.\n
Boundary markers: The references/base-prompt.md template lacks robust delimiters or explicit instructions to ignore embedded commands within the {{CONTENT}} block.\n
Capability inventory: The skill writes multiple files to the local file system (analysis.md, structured-content.md, prompts/infographic.md), performs shell checks, and triggers external image generation capabilities.\n
Sanitization: The skill instructions explicitly require the agent to preserve source data verbatim, preventing any sanitization or filtering of potentially malicious instructions in the input.\n- [COMMAND_EXECUTION]: The skill executes shell commands (test -f) to verify the existence of optional extension configuration files (EXTEND.md) in the current project and the user's home directory.

baoyu-infographic