The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local helper script (scripts/wildcard.py) using python3 to generate randomized creative directions. It also uses curl -L to download images from URLs discovered during the research phase.
[EXTERNAL_DOWNLOADS]: Fetches visual data from well-known and reputable services including The Metropolitan Museum of Art, Openverse, and Are.na. These operations are core to the skill's purpose of building moodboards from public domain and open-access sources.
[INDIRECT_PROMPT_INJECTION]: The skill parses metadata.prompt fields from public assets and description fields from external APIs to "research wording."
Ingestion points: External text metadata from the Scenario public catalog and third-party APIs (Are.na, The Met, etc.).
Boundary markers: The instructions do not define explicit delimiters to prevent the agent from being influenced by instructions embedded within the researched metadata.
Capability inventory: Shell execution (curl), local script execution (python3), and the ability to create/update collections and upload assets.
Sanitization: No explicit sanitization or filtering of external prompt strings is implemented before the agent interprets them for style research.

scenario-inspiration