The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown-based instructions that guide the AI agent's analysis process. It contains no executable binaries, scripts, or commands that modify the environment or persistence.- [EXTERNAL_DOWNLOADS]: The skill references a static image from Alibaba's CDN (alicdn.com) for instructional layout purposes and mentions the author's GitHub repository. Both are well-known or vendor-owned sources and do not involve script execution or sensitive data harvesting.- [INDIRECT_PROMPT_INJECTION]: The skill captures user-provided strings for evaluation as prompts. (1) Ingestion points: User input captured via the /prompt command or extracted from conversation (SKILL.md). (2) Boundary markers: The skill instructs the agent to treat specific text segments as the analysis target but does not include explicit escaping for instructions contained within that text. (3) Capability inventory: The skill has no capabilities to execute code, write to the filesystem, or make network requests. (4) Sanitization: None. Because the skill lacks sensitive tools, the ingestion of untrusted prompt text for analysis is safe.

prompt