The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several operational scripts intended for local execution by the user or agent, including check_services.mjs, sync_domains.mjs, and ingest_spend.mjs. These scripts are used to perform health checks and manage the data layer in Busabase.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes data from external sources.
Ingestion points: External data is ingested from user-defined service URLs in scripts/check_services.mjs and from public RDAP records via https://rdap.org in scripts/sync_domains.mjs.
Boundary markers: The system requires human approval for all agent-proposed action cards (e.g., domain renewals, service restarts) within the AirApp UI, preventing the agent from acting autonomously on potentially poisoned data.
Capability inventory: The scripts perform network probes (HTTP/TLS/RDAP) and write data to Busabase. Critical operations like service restarts are performed by the agent in the external environment only after user approval.
Sanitization: The implementation includes HTML escaping functions (escapeHtml) in both the UI logic and the data models to prevent cross-site scripting (XSS) when rendering ingested data.

kelly-devops