Problem
rtk is a Claude Code-specific CLI proxy that filters command output to reduce token usage (60-90% savings). No equivalent exists in Gemini CLI, GitHub Copilot, or OpenAI Codex.
Currently /convert marks this as unsupported with no guidance.
Proposed Solution
Document tool-specific context management strategies in each generated config file:
- Gemini CLI:
model.summarizeToolOutput in .gemini/settings.json can cap token output per tool (e.g., limit run_shell_command output to N tokens)
- Copilot: Instruct users to use
# file references instead of pasting full file contents
- Codex: Document context window best practices for large repos
Acceptance Criteria
Ref: #11
Problem
rtkis a Claude Code-specific CLI proxy that filters command output to reduce token usage (60-90% savings). No equivalent exists in Gemini CLI, GitHub Copilot, or OpenAI Codex.Currently
/convertmarks this as unsupported with no guidance.Proposed Solution
Document tool-specific context management strategies in each generated config file:
model.summarizeToolOutputin.gemini/settings.jsoncan cap token output per tool (e.g., limitrun_shell_commandoutput to N tokens)#file references instead of pasting full file contentsAcceptance Criteria
summarizeToolOutputconfig exampleRef: #11