The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use browser or image tools to inspect media hosted on external domains, specifically goodcase.ai and media.goodcase.ai. While these links are contextually relevant to the video prompt engineering task, they involve interaction with external infrastructure not listed as trusted.
[INDIRECT_PROMPT_INJECTION]: The workflow involves reading and processing "Prompt excerpts" and "Case evidence" from references/cases.md. This file contains structured text that the agent is expected to synthesize into new artifacts. This represents a potential surface for indirect prompt injection if the external data contained hidden instructions, though no such malicious content was found in the provided files.
[COMMAND_EXECUTION]: The skill mentions turning the method into an "executable, Case-grounded workflow," but analysis shows this refers to a logical sequence of operations for the agent to follow rather than the execution of system commands or scripts.

retro-dv-home-video-by-wat8s8