The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external content (images and web pages) from third-party sources like goodcase.ai and x.com using browser or image inspection tools. This process creates an attack surface where malicious instructions could be embedded in the external content to influence the agent's behavior during the workflow.
Ingestion points: The agent is directed to read and inspect media from multiple external URLs listed in the references/cases.md file.
Boundary markers: The instructions define a structured 'Preserve / Replace / Avoid' contract, but do not include explicit instructions to ignore potentially malicious commands embedded within the retrieved external data.
Capability inventory: The skill utilizes standard browser and image tools. It does not include commands for local file writing, shell execution, or sensitive data exfiltration.
Sanitization: There is no evidence of sanitization or filtering applied to the content retrieved from external sources.
[EXTERNAL_DOWNLOADS]: The skill references and directs the agent to browse external websites to gather visual evidence and design patterns. These include links to an industry-specific case study site (goodcase.ai) and a well-known social media platform (x.com). These references are core to the skill's primary purpose of visual reinterpretation and do not involve the installation of external software packages or scripts.

photo-reinterpretation-diptych