The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The file references/cases.md contains multiple URLs to goodcase.ai and x.com. The skill's workflow instructs the agent to inspect these external media files and summaries to ground the generated prompts in specific evidence. These downloads are relevant to the skill's primary purpose and involve well-known services.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (GoodCase summaries and media content), which creates an attack surface for indirect prompt injection. However, this is a standard capability for this type of skill, and the instructions include a 'Verification' section and 'Safety and attribution' rules that require the agent to distinguish observed evidence from its own recommendations, providing a layer of protection against accidental instruction following from processed data.

pov-vlog-presence-by-1sduwzc