suppressing-noisy-errors
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes error messages, exception types, and stack traces originating from external sources, creating a surface for indirect prompt injection where malicious instructions could be embedded in error logs.
- Ingestion points: Untrusted error data is ingested via
posthog:query-error-tracking-issues-listandposthog:query-error-tracking-issue-eventsas described in Step 1 and Step 2 of the workflow. - Boundary markers: The instructions do not define specific delimiters to wrap the interpolated error data, which could lead the agent to interpret data as instructions.
- Capability inventory: The skill has the ability to execute SQL queries via
posthog:execute-sqland modify ingestion behavior viaposthog:error-tracking-suppression-rules-create. - Sanitization: The workflow mitigates this risk by requiring a dry-run preview of the filter's impact (Step 5) and explicit human confirmation (Step 6) before any permanent changes are made.
- [COMMAND_EXECUTION]: The skill utilizes the
posthog:execute-sqltool to perform dry-runs of suppression filters against live data when standard filtering logic is insufficient. This tool allows the agent to construct and execute SQL/HogQL queries within the PostHog environment to validate rule accuracy.
Audit Metadata