dt-obs-log-semantic-mapping
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is authored by 'dynatrace' and operates exclusively on vendor-specific log formats and semantic dictionary definitions consistent with Dynatrace documentation.
- [COMMAND_EXECUTION]: The skill includes instructions for running DQL (Dynatrace Query Language) commands. These are scoped to the intended purpose of fetching log samples from the user's tenant for mapping validation and do not demonstrate command injection or privilege escalation patterns.
- [EXTERNAL_DOWNLOADS]: The skill references the official Dynatrace documentation for its Log Semantic Dictionary. This is a well-known service from a trusted source and does not involve downloading executable code.
- [DATA_EXFILTRATION]: The skill analyzes log data but does not contain patterns for exfiltrating sensitive information to external domains. It focuses on internal data mapping within the Dynatrace ecosystem.
- [PROMPT_INJECTION]: No evidence of prompt injection or instructions to bypass safety guardrails was found in the skill body or metadata.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external log data (Category 8).
- Ingestion points: Raw vendor logs are ingested via manual paste or live tenant queries (
references/runtime-validation.md). - Boundary markers: The instructions specify deserializing the
contentstring into structured JSON, which helps maintain data boundaries. - Capability inventory: Limited to DQL execution and text analysis within the agent environment.
- Sanitization: The skill recommends mapping to strict semantic dictionary enums and types (e.g.,
loglevel,status), which acts as a form of validation/sanitization. Severity is LOW.
Audit Metadata