sentry-setup-ai-monitoring
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses standard environment detection commands (
grep) to identify installed AI SDKs in project configuration files such aspackage.json,requirements.txt, andcomposer.json.- [SAFE]: The skill provides explicit warnings about capturing user content (PII) and instructs the agent to seek user confirmation before enabling prompt or output recording, following security best practices.- [SAFE]: Configuration samples use standard Sentry SDK patterns and placeholders for sensitive information like DSNs, avoiding the inclusion of hardcoded credentials.- [INDIRECT_PROMPT_INJECTION]: The skill configures a pipeline that processes untrusted user prompts and agent outputs, creating an attack surface for indirect prompt injection if the monitored data contains malicious instructions. - Ingestion points: The skill reads local project metadata (
package.json,requirements.txt,composer.json) to determine the SDK environment. - Boundary markers: The skill includes a 'Data Capture Warning' section that explicitly highlights the risk of PII exposure and requires user consent.
- Capability inventory: The skill uses
grepfor file discovery and provides code templates for initializing Sentry monitoring, which involves capturing and transmitting LLM transaction data. - Sanitization: The skill relies on the underlying Sentry SDK's default data scrubbing and requires explicit user opt-in for capturing prompt/response content.
Audit Metadata