The Agent Skills Directory

[DATA_EXFILTRATION]: The skill provides a TypeScript code template in Step 3 for a custom VS Code extension that collects and transmits telemetry. The template specifically captures vscode.env.machineId and doc.fileName (the path of every file saved by the user) and sends this data to a remote network endpoint via the fetch API. Although the destination endpoint is read from a user's configuration, the template facilitates the automated transmission of local file metadata and system identifiers to an external server.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for processing untrusted data from external analytics dashboards and custom telemetry events, which represents a potential attack surface.
Ingestion points: Analytical data from the Windsurf Admin Dashboard and usage events captured by the custom extension described in SKILL.md.
Boundary markers: The instructions lack specific delimiters or instructions for the agent to ignore potentially malicious content embedded within the ingested analytics data.
Capability inventory: The skill has access to Read, Write, and Edit tools, which could be misused if instructions are injected via the metrics being analyzed.
Sanitization: There is no evidence of sanitization or validation of the external metrics before they are used to generate productivity reports or dashboard visualizations.

windsurf-observability