The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for the agent to process external data, including product feature descriptions and website URLs (interpolated into the agent-browser tool). This ingestion of untrusted content creates a surface for indirect prompt injection, where an attacker could embed malicious instructions in the content being summarized or captured. The provided templates do not currently include explicit boundary markers or instructions to ignore embedded commands within the processed data.
Ingestion points: User-provided feature descriptions and URLs processed in SKILL.md via tools like agent-browser.
Boundary markers: Absent from the recommended changelog and visual generation templates.
Capability inventory: The skill uses the Bash(belt *) tool to execute various remote AI applications, including a browser for screenshots and image generation models.
Sanitization: No specific sanitization or filtering logic is prescribed for the external inputs.
[EXTERNAL_DOWNLOADS]: The skill documentation references and facilitates the installation of the belt CLI and related skill packages from the inference-sh and belt-sh organizations. These are external dependencies required for the skill's visual generation features.

product-changelog