The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to run several Node.js scripts in the skill's root directory to perform its primary functions, such as scripts/ingest_signals.mjs for data entry, scripts/file_report.mjs for research filing, and scripts/execute_decisions.mjs for handoff operations.
[EXTERNAL_DOWNLOADS]: The skill uses standard Node.js package managers and depends on well-known libraries, including busabase-sdk for database communication and hono for the local preview server.
[PROMPT_INJECTION]: The skill processes external market data, changelogs, and news signals, creating a potential surface for indirect prompt injection.
Ingestion points: Data is ingested via scripts like ingest_signals.mjs and ingest_trends.mjs from monitored competitor URLs and search results.
Boundary markers: The instructions mandate respect for robots.txt and site terms, focusing strictly on public information.
Capability inventory: The skill can execute local scripts, write to the Busabase database via API, and store local JSON payloads.
Sanitization: The dashboard UI uses escapeHtml to prevent XSS when displaying ingested content, and all sensitive actions (like handoffs to other skills) require manual user approval within the app.

kelly-radar