authoring-log-alerts
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled local Python script (
scripts/baseline_stats.py) and Python one-liners via the shell. These scripts perform mathematical calculations on log frequency data to suggest appropriate alert thresholds. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect injection surface as it processes untrusted log data to derive alert configurations.
- Ingestion points: Data is ingested through the
posthog:logs-count-rangestool (SKILL.md). - Boundary markers: The instructions mandate the use of
posthog:logs-alerts-simulate-createto back-test and validate alert logic against historical data before deployment. - Capability inventory: Includes shell execution for script processing and several PostHog API tools for log analysis and alert management.
- Sanitization: The included Python script uses standard JSON parsing to process input data, reducing the risk of shell injection or malformed data issues.
Audit Metadata