The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/model-error-wrapper.sh executes shell commands provided via the --command argument using bash -c to implement automatic retries and error monitoring for model requests.\n- [COMMAND_EXECUTION]: The scripts/auto-switch-handler.sh automation handler invokes local scripts and standard Python 3 utilities for state management and model switching logic.\n- [DATA_EXFILTRATION]: The skill documentation includes examples for configuring Telegram notifications via curl. This utilizes a well-known service for operational alerting and does not involve the transfer of sensitive data.\n- [SAFE]: The skill processes command output to detect failure strings, creating a surface for indirect prompt injection. Ingestion points: scripts/model-error-wrapper.sh reads log file tails. Boundary markers: Absent. Capability inventory: Executes scripts/auto-switch-handler.sh to transition model configurations. Sanitization: Absent. This surface is limited to triggering intended failover behaviors.

model-fallback