right now the model is provided to the engine directly --model gpt-5.3-codex :
copilot --add-dir /tmp/ --add-dir /tmp/gh-aw/ --add-dir /tmp/gh-aw/agent/ --log-level all --log-dir /tmp/gh-aw/sandbox/agent/logs/ --disable-builtin-mcps --model gpt-5.3-codex --allow-tool 'shell(cat)' --allow-tool 'shell(grep)' --allow-tool 'shell(head)' --allow-tool 'shell(jq)' --allow-tool 'shell(ls)' --allow-tool 'shell(tail)' --allow-tool 'shell(wc)' --share /tmp/gh-aw/sandbox/agent/logs/conversation.md --prompt "$COPILOT_CLI_INSTRUCTION" 2>&1 | tee /tmp/gh-aw/threat-detection/detection.log
Ideally the model would be used to populate an environment variable and then the env var used in the call.
Otherwise if you try to make the model a variable or a workflow_call input, the variable syntax ends up in the shell command directly and the compiler refuses to compile it (due to prompt injection validation)
Copilot doesnt accept the model as an env var directly, so you'd still need to reference the env var in the shell command.
right now the model is provided to the engine directly
--model gpt-5.3-codex:Ideally the model would be used to populate an environment variable and then the env var used in the call.
Otherwise if you try to make the model a variable or a workflow_call input, the variable syntax ends up in the shell command directly and the compiler refuses to compile it (due to prompt injection validation)
Copilot doesnt accept the model as an env var directly, so you'd still need to reference the env var in the shell command.