The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and act upon data from external task briefs and instructions to guide its decision-making logic for image adaptation. This interpolation of untrusted data creates a surface where instructions embedded in the briefs could influence the agent's execution flow.
Ingestion points: The agent is directed to use the 'task's named master', 'brief's own description', and 'task instructions' (SKILL.md) to determine rungs on the decision ladder.
Boundary markers: No specific delimiters or safety warnings are provided to separate these external inputs from the agent's core instructions.
Capability inventory: The skill utilizes model_run for generative operations, search for discovering models, and asset_analyze for validation, all of which are parameterized by the input data.
Sanitization: There are no defined processes for sanitizing or filtering the content of the task briefs before they are processed by the agent.
[COMMAND_EXECUTION]: The instructions guide the agent to recommend the command npx skills add scenario-labs/skills --skill <name> to the user for installing missing dependencies. This involves the execution of external packages via a shell command. The reference targets the official repository of the skill author and is documented as a standard setup and recovery procedure for this environment.

scenario-formats