error-diagnostics-error-trace
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill contains logic to read and analyze source code files, creating a potential surface for indirect prompt injection.
- Ingestion points: The
ErrorTrackingAnalyzerclass inresources/implementation-playbook.mdreads content from files within a target project path. - Boundary markers: The agent is instructed to stop for clarification if safety boundaries or permissions are missing, and to define a redacted schema for events.
- Capability inventory: The skill allows the agent to perform file system analysis and communicate results to external monitoring services.
- Sanitization: Implementation examples include a data sanitization routine that redacts common sensitive keys such as passwords and API tokens.
- [DATA_EXFILTRATION]: The skill enables sending monitoring data to external endpoints and webhooks.
- Evidence: Code templates in
resources/implementation-playbook.mddemonstrate sending telemetry to external endpoints (ErrorTracker) and Slack webhooks (SlackAlertChannel). These features are designed for observability but require user-provided credentials and endpoints.
Audit Metadata