The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/model-error-wrapper.sh executes shell commands via the --command parameter using bash -c. This behavior is essential for its intended function as a wrapper to manage retries and error handling for model requests.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it captures and processes error messages from the output of wrapped commands.\n
Ingestion points: scripts/model-error-wrapper.sh captures the output of executed commands and passes it to scripts/auto-switch-handler.sh for error analysis.\n
Boundary markers: No specific delimiters or instructions to ignore embedded commands are used when processing or logging captured error messages.\n
Capability inventory: The skill executes local scripts and shell commands as part of its model management logic.\n
Sanitization: Error messages are captured from command output and logged without sanitization or validation.

model-fallback