The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved via Exa searches and Firecrawl crawls/scrapes.
Ingestion points: Data enters the agent's context through monitor.runs, monitor.site.checks, and raw results fetched from external websites.
Boundary markers: The skill does not explicitly define delimiters for external content, though it instructs the agent to summarize results rather than outputting raw payloads.
Capability inventory: The skill possesses the ability to execute shell commands (via the arks CLI) and manage network-accessible webhooks.
Sanitization: The agent is instructed to "Summarize run or check history instead of pasting raw Provider payloads," which provides a level of interpretation but does not fully mitigate risks from instructions embedded in the web content.
[COMMAND_EXECUTION]: The skill relies on executing the arks CLI tool on the local host to manage monitors.
Evidence: The instructions direct the agent to execute various shell commands, including arks doctor --json and arks monitor site create --input ./site-monitor.json, with arguments derived from user-supplied queries, schedules, and targets.

monitor