bug: Gemini consultation fails — tries to use non-existent run_shell_command tool #652
Description
Bug
When running consult -m gemini for code review, Gemini repeatedly fails to produce output because it tries to use run_shell_command which doesn't exist in its agent environment.
Context
This is happening during SPIR protocol implementation reviews. The consultation is supposed to review code diffs, not execute shell commands. Gemini's agent may be misconfigured or defaulting to a tool that isn't available.
Expected
Gemini should review the code using file access tools (read, glob, grep) and produce a verdict, not try to run shell commands.
Environment
Shannon project using codev SPIR protocol with consult -m gemini --protocol spir --type impl