[oharboe]

ORFS has no Gemini review guidance. Gemini review comments cost authors time to refute, and the most common advice that gets turned down is also the most harmful when followed: guarding a variable that variables.yaml already defaults, or a fallback that turns a missing file or a malformed setting into an empty result, so an hours-long run finishes green and measures nothing. This adds .gemini/styleguide.md with that as a hard stop, points at the Tcl and Makefile contracts instead of repeating them, and lists the other false positives that recur.

Where the rules come from

Every Gemini review thread on ORFS pull requests since April 2026 that got a human reply (190 of 641 comments) was read and the turned-down advice sorted by habit:

Habit	Threads	Examples
Defensive guard, default or swallowed error where a hard early failure is wanted	26	#4563, #4543, #4597, #4660, #4389, #4744
Wrong fact about a tool, PDK, package or the repo (yosys clockgate -liberty, "V9 does not exist", write_lec_verilog paths, raised twice)	18	#4425, #4364, #4365
Misreads make, shell, Tcl or yosys select semantics ($$$$ in eval, empty for list)	10	#4596, #4649, #4175, #4524
Edits to code moved verbatim, vendored PDK files, one-off cleanups	12	#4603, #4277, #4609
Lectures on rules-base.json, QoR or constraint choices the PR already explains	12
Style nits against local convention	12	#4353, #4179
Portability ORFS does not support (RHEL 7, paths with spaces)	4	#4553, #4555, #4556

The style guide is read by Gemini Code Assist from .gemini/styleguide.md without a config.yaml; review settings are left as they are.

🤖 Generated with Claude Code

[gemini-code-assist]

Note

Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported.