The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands via bash to manage temporary data files in /tmp/earnings-preview/ and to display the final report using the open command. This is standard behavior for the skill's documented purpose of managing large data sets and providing a rendered report.
[EXTERNAL_DOWNLOADS]: The HTML report template loads the Chart.js and chartjs-plugin-annotation libraries from the well-known jsDelivr CDN to support data visualization. This is a legitimate use of an established third-party service.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from earnings transcripts and news search results.
Ingestion points: Untrusted content enters the agent's context through get_transcript_from_key_dev_id and the Kensho search tool.
Boundary markers: The skill employs structural delimiters such as HTML <blockquote> tags for verbatim quotes and specific <a> tags for data references, providing some separation between narrative and external data.
Capability inventory: The agent has file system access (mkdir, cat) and the ability to execute the generated report via open.
Sanitization: The instructions do not specify programmatic sanitization or escaping for the external text before it is interpolated into the HTML/JavaScript report.
[REMOTE_CODE_EXECUTION]: The skill performs dynamic execution by generating an HTML file with embedded JavaScript logic to render interactive charts. While the scripts follow a static template, the data used by the JavaScript is sourced from external MCP tool outputs.

earnings-preview-single