What problem does it solve?
Autonomous security agents often get stuck retrying the same failing command with minor parameter tweaks, wasting tool calls without making progress. This Skill detects stagnation patterns and forces a structured diagnosis-and-recovery workflow so the agent changes direction instead of spinning in place.
Core Features & Use Cases
- Stagnation Diagnosis: Classifies the stuck state into five modes (looping on one approach, environment constraints, insufficient reconnaissance, broken attack chains, or tool misuse) before taking action.
- Attack Tree Redirection: Provides Web, privilege escalation, and cloud (Azure/AWS/GCP) attack trees so the agent picks a genuinely new direction rather than a parameter variation.
- Script Substitution: Replaces error-prone curl commands and multi-step API flows with Python scripts to avoid shell escaping and state-passing failures.
- Use Case: During an authorized penetration test or CTF, the agent receives repeated 403 errors on an OAuth endpoint; the Skill guides it to stop retrying, enumerate untried attack surfaces, and pivot from client_credentials to a device_code flow.
Quick Start
Ask the agent to apply the stagnation-recovery procedure after it has made no progress for ten or more tool calls on the current security task.
