grouping-noisy-errors
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes $exception events, including error messages and stack traces, which originate from external applications. These data fields could be used to host malicious instructions intended to trick the agent into incorrectly merging issues or creating broad grouping rules that hide legitimate errors.
- Ingestion points: Error tracking data is fetched via
posthog:query-error-tracking-issue-eventsin SKILL.md. - Boundary markers: The skill documentation includes an explicit "Are they the same error?" checklist and a list of "What NOT to group together" to ensure the agent performs strict verification before taking action.
- Capability inventory: The skill has the ability to merge issues (
posthog:error-tracking-issues-merge-create), create grouping rules (posthog:error-tracking-grouping-rules-create), and update issue metadata (posthog:error-tracking-issues-partial-update). - Sanitization: The skill relies on specific property comparison (e.g., matching $lib, exception types, and stack frames) rather than content sanitization to validate incoming data.
Audit Metadata