The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to rely on potentially untrusted "task instructions" or a "brief's own description" as authoritative guides, especially when operating unattended. This creates an attack surface where an attacker could embed instructions in a project brief to manipulate how the agent uses image tools.
Ingestion points: SKILL.md explicitly directs the agent to use "the task's named master or the brief's own description" and "task instructions" for decision-making.
Boundary markers: No delimiters or safety instructions are used to distinguish user instructions from data provided in the brief.
Capability inventory: The skill utilizes model_run, search, asset_analyze, and asset_display tools across its operations.
Sanitization: No evidence of sanitization or validation of the input description is present before it influences the model's decision ladder.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to recommend the installation of additional packages from the author's official repository.
Evidence: SKILL.md contains the pattern: "ask the user to install it (npx skills add scenario-labs/skills --skill <name>)."
Context: The package scenario-labs/skills belongs to the official vendor scenario-labs and is presented as a legitimate extension.

scenario-formats