log-analytics
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external log data from OpenSearch indices, exposing it to indirect prompt injection vulnerabilities.
- Ingestion points: The skill reads log data via
SearchIndexTool,SearchIndexTool, andGenericOpenSearchApiToolas described inSKILL.mdandlog-analytics.md. - Boundary markers: No specialized delimiters or instruction guarding are used to isolate untrusted log contents from system instructions.
- Capability inventory: The toolset includes running PPL queries, searching the web via
ddg-search, and interacting with the OpenSearch API. - Sanitization: The instructions do not define any sanitization, escaping, or verification of the content inside the log messages before processing.
- [CREDENTIALS_UNSAFE]: In
log-analytics.md, a hardcoded default/mock credential string (My_password_123!@#) is specified in the fallback configuration table for theOPENSEARCH_PASSWORDenvironment variable. - [COMMAND_EXECUTION]: In
log-analytics.md, the documentation instructs the agent to execute a script pathscripts/opensearch_ops.pyviauv run pythonwhich is not delivered within the skill files bundle.
Audit Metadata