The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data, such as webhook payloads and Uniform entry content, using AI agents, which introduces a risk surface for instructions embedded in that data.
Ingestion points: The Scout agent and ScoutClient ingest content from Uniform entries and incoming webhook payloads as context for AI tasks (identified in templates/scout-workflow-stage.automation.ts and templates/scout-client.automation.ts).
Boundary markers: The documentation recommends using structured output schemas (Zod) and specific instructional framing to bound agent behavior, but the templates do not implement strict input validation to detect or strip adversarial instructions from processed content.
Capability inventory: The skill possesses capabilities to modify Uniform entries via EntryManagementClient.save, transition workflow stages via WorkflowClient, and interact with external notification services like Slack (identified in references/sdk-api.md and references/scout.md).
Sanitization: The skill relies on natural language constraints ("The bar for rejection is high", "Do not invent product claims") and platform-level AI guardrails rather than programmatic sanitization of ingested content.

uniform-automations