investigating-with-observability
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external observability backends, which presents a surface for indirect prompt injection attacks.
- Ingestion points: Untrusted data enters the agent context through the outputs of metrics, logs, and traces APIs fetched by subagents (defined in
agents/metrics-discovery.md,agents/logs-discovery.md,agents/traces-discovery.md, andagents/alertmanager-check.md). - Boundary markers: The skill lacks explicit boundary markers or instructions that warn the agent to ignore potentially malicious instructions embedded within the ingested telemetry data.
- Capability inventory: The skill possesses the capability to execute shell commands (
curl,jq,date) and dispatch further subagents based on its analysis of the ingested data. - Sanitization: There is no evidence of content sanitization or escaping of API responses before they are processed by the LLM to form hypotheses or next steps.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute
curlcommands to interact with VictoriaMetrics APIs. These commands are well-structured and use environment variables for dynamic configuration. Notably, the skill follows secure patterns by using a configuration file for authentication headers (VM_CURL_CONFIG) to avoid exposing credentials in command-line arguments.
Audit Metadata