The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple external sources which can lead to indirect prompt injection vulnerabilities.
Ingestion points: Data is ingested from external sources via himalaya (email), google-workspace (email/calendar), web_extract/browser_navigate (web content), and read_file (local filesystem) as specified in Procedures 2 and 5.
Boundary markers: The instructions lack specific guidance on using delimiters or "ignore embedded instructions" markers when the agent processes retrieved content.
Capability inventory: The skill possesses capabilities to write to the filesystem (dashboard.json, index.html), schedule recurring tasks via the cronjob tool, and open files/URLs via desktop_preview.
Sanitization: There is no mention of sanitizing, escaping, or validating the external content before it is interpolated into the dashboard state or summaries.
[PERSISTENCE]: The skill utilizes a persistence mechanism to maintain recurring functionality.
Evidence: Procedure 4 explicitly instructs the agent to use the cronjob tool to create a per-dashboard job for recurring refresh ticks. This is the primary intended purpose of the skill to ensure dashboards remain updated without manual intervention.

live-dashboard