Composable detection and maintenance skill prompts for AI security workflows
Log Anomaly Analyzer
Ingests a log sample and a baseline profile, scores each event for anomaly likelihood, and returns ranked findings with suggested detection rule titles.
Goal Drift Detector
Analyzes an AI agent's action log against its declared objective to detect gradual goal misalignment, scope expansion, or rogue behavior emergence.
Prompt Injection Detector
Scans agent input text for embedded instruction injection patterns, semantic override attempts, and formatting-based jailbreaks. Returns injection probability and the offending substring.
Tool Misuse Detector
Compares observed agent tool calls against an authorized tool manifest and resource scope policy. Flags destructive verbs, out-of-scope targets, and call rate anomalies.
Data Exfiltration Detector
Correlates network flow logs and file events attributed to agent processes to detect large-volume transfers, encoding-based staging, and connections to unapproved external destinations.
Detection Generator
Takes a threat description and optional sample log, then generates production-ready Sigma, Splunk SPL, and PySpark detection rules with field mappings, false positive guidance, and severity classification.
Select a skill to inspect
Choose from the panel on the left