The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing official SDK packages such as @sentry/node, @sentry/bun, and @sentry/deno via standard package managers (npm, yarn, pnpm, bun). It also suggests using the @sentry/wizard via npx for interactive setup. These are trusted resources belonging to the skill's author (getsentry).
[COMMAND_EXECUTION]: The detection phase uses standard shell commands (bun --version, node --version, cat package.json, ls, grep) to identify the project environment. These commands are limited to local discovery and are non-malicious in the context of SDK configuration.
[INDIRECT_PROMPT_INJECTION]: The skill documents AI monitoring features that capture LLM inputs and outputs. While this represents a potential data ingestion surface, the SDK includes built-in privacy controls (dataCollection.genAI) to redact or disable prompt capture, which are clearly documented for the user.
Ingestion points: The skill facilitates monitoring of AI libraries (OpenAI, Anthropic, etc.), where LLM prompts/responses enter the agent's telemetry context.
Boundary markers: The SDK uses internal span management for delimitation; instructions for user-defined boundary markers are not present as the instrumentation is largely automatic.
Capability inventory: The SDK is configured to perform network operations to send telemetry data to the vendor's ingest servers (*.sentry.io).
Sanitization: The documentation provides clear guidance on using the dataCollection.genAI toggle to disable or redact sensitive AI inputs and outputs.

sentry-node-sdk