fix: switch multi-model frontend routing from Gemini CLI to Antigravity CLI#2520
fix: switch multi-model frontend routing from Gemini CLI to Antigravity CLI#2520vitalii-novikov wants to merge 1 commit into
Conversation
…ty CLI Google is sunsetting consumer Gemini CLI access on 2026-06-18 and consolidating into Antigravity CLI. codeagent-wrapper already ships a working AntigravityBackend (confirmed by shelling to `agy`), and ~/.claude/.ccg/prompts/antigravity/*.md role prompts already exist — only the command markdown files still hardcoded --backend gemini. Updates multi-frontend.md, multi-execute.md, multi-plan.md, and multi-workflow.md to route frontend calls through --backend antigravity instead of --backend gemini, point role-prompt paths at prompts/antigravity/ instead of prompts/gemini/, and drop the gemini-only --gemini-model flag (antigravity has no CLI equivalent; codeagent-wrapper picks its default model). multi-backend.md is unaffected (codex-only, no frontend routing).
|
ECC bundle files are already tracked in this repository. Skipping generation of another bundle PR. |
📝 WalkthroughSummary by CodeRabbit
WalkthroughThe command documentation replaces Gemini with Antigravity as the frontend authority across model invocation, role prompts, session reuse, routing, planning, execution, review, and trust guidance. ChangesAntigravity workflow migration
Estimated code review effort: 2 (Simple) | ~10 minutes Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
| Filename | Overview |
|---|---|
| commands/multi-frontend.md | Switches the frontend-led workflow to Antigravity calls, prompt paths, and session names; the first call now depends on a compatible external wrapper and prompt install. |
| commands/multi-plan.md | Updates planning instructions, prompt paths, and plan handoff fields from Gemini to Antigravity. |
| commands/multi-execute.md | Updates execution routing, implementation prompts, review prompts, and trust rules to use Antigravity for frontend work. |
| commands/multi-workflow.md | Updates the full multi-model workflow to use Antigravity for frontend analysis, planning, and review. |
Reviews (1): Last reviewed commit: "fix: switch multi-model frontend routing..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| # New session call | ||
| Bash({ | ||
| command: "~/.claude/bin/codeagent-wrapper {{LITE_MODE_FLAG}}--backend gemini --gemini-model gemini-3-pro-preview - \"$PWD\" <<'EOF' | ||
| command: "~/.claude/bin/codeagent-wrapper {{LITE_MODE_FLAG}}--backend antigravity - \"$PWD\" <<'EOF' |
There was a problem hiding this comment.
When a user runs this updated workflow with the ccg-workflow runtime named in the prerequisite, this first Antigravity call now depends on codeagent-wrapper accepting --backend antigravity and having matching Antigravity role prompts. If their provisioned wrapper or prompt tree still exposes the old Gemini backend layout, the command fails before Phase 2 can produce a session; the command should state the required ccg-workflow/codeagent-wrapper version or reprovision step at the migration point.
Rule Used: Review for prompt injection, tool-permission creep... (source)
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
commands/multi-execute.md (1)
217-237: 🔒 Security & Privacy | 🟠 Major | 🏗️ Heavy liftAdd one explicit untrusted-model boundary across all Antigravity workflows.
The migration consistently elevates Antigravity output, but Claude ultimately applies model-generated plans and diffs. Validate approved scope, file paths, tool/shell instructions, and secrets before any write or downstream handoff.
commands/multi-execute.md#L217-L237: validate unified diffs before Edit/Write.commands/multi-workflow.md#L32-L34: describe specialization without unconditional trust.commands/multi-plan.md#L149-L153: validate cross-model planning output against the requirement.commands/multi-plan.md#L272-L274: make session handoff data subject to the same validation.commands/multi-frontend.md#L27-L29: remove unconditional “trustworthy” wording.commands/multi-frontend.md#L161-L164: apply the same trust boundary to frontend review output.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@commands/multi-execute.md` around lines 217 - 237, Establish an explicit untrusted-model validation boundary across all Antigravity workflows: in commands/multi-execute.md lines 217-237, validate approved scope, paths, tool/shell instructions, secrets, and unified diffs before Edit/Write or handoff; in commands/multi-workflow.md lines 32-34 and commands/multi-frontend.md lines 27-29, remove unconditional trust language and describe specialization as untrusted output; in commands/multi-plan.md lines 149-153 and 272-274, validate planning and session-handoff data against the requirement; and in commands/multi-frontend.md lines 161-164, apply the same validation boundary to frontend review output.Source: Path instructions
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@commands/multi-execute.md`:
- Around line 217-237: Establish an explicit untrusted-model validation boundary
across all Antigravity workflows: in commands/multi-execute.md lines 217-237,
validate approved scope, paths, tool/shell instructions, secrets, and unified
diffs before Edit/Write or handoff; in commands/multi-workflow.md lines 32-34
and commands/multi-frontend.md lines 27-29, remove unconditional trust language
and describe specialization as untrusted output; in commands/multi-plan.md lines
149-153 and 272-274, validate planning and session-handoff data against the
requirement; and in commands/multi-frontend.md lines 161-164, apply the same
validation boundary to frontend review output.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 0b60fb2d-7767-4ebc-a496-1206266b1cbc
📒 Files selected for processing (4)
commands/multi-execute.mdcommands/multi-frontend.mdcommands/multi-plan.mdcommands/multi-workflow.md
📜 Review details
⏰ Context from checks skipped due to timeout. (1)
- GitHub Check: Greptile Review
🧰 Additional context used
📓 Path-based instructions (4)
commands/**/*.md
📄 CodeRabbit inference engine (CLAUDE.md)
Commands should be formatted as Markdown with description frontmatter.
Files:
commands/multi-frontend.mdcommands/multi-workflow.mdcommands/multi-plan.mdcommands/multi-execute.md
{agents,skills,commands}/**/*.md
📄 CodeRabbit inference engine (CLAUDE.md)
Use lowercase filenames with hyphens (e.g.,
python-reviewer.md,tdd-workflow.md) for agents, skills, and commands.
Files:
commands/multi-frontend.mdcommands/multi-workflow.mdcommands/multi-plan.mdcommands/multi-execute.md
commands/**
📄 CodeRabbit inference engine (AGENTS.md)
commands/is a legacy slash-entry compatibility surface and should only be added or updated when a shim is still required for migration or cross-harness parity.
Files:
commands/multi-frontend.mdcommands/multi-workflow.mdcommands/multi-plan.mdcommands/multi-execute.md
{skills,commands,agents,rules}/**
⚙️ CodeRabbit configuration file
{skills,commands,agents,rules}/**: Focus on prompt-injection resilience, tool-permission scope, destructive action guards, and secret exfiltration risks.
Files:
commands/multi-frontend.mdcommands/multi-workflow.mdcommands/multi-plan.mdcommands/multi-execute.md
🔇 Additional comments (4)
commands/multi-workflow.md (1)
7-7: LGTM!Also applies to: 23-23, 45-45, 60-60, 75-83, 128-128, 140-142, 154-160, 176-176
commands/multi-plan.md (1)
18-18: LGTM!Also applies to: 31-31, 46-46, 48-53, 131-131, 140-145, 164-165, 178-183, 201-201
commands/multi-execute.md (1)
19-19: LGTM!Also applies to: 32-32, 47-47, 65-65, 87-94, 135-137, 180-183, 202-202, 251-261, 284-286, 298-299
commands/multi-frontend.md (1)
7-7: LGTM!Also applies to: 20-21, 41-41, 56-56, 72-78, 94-94, 105-122, 139-145
daltino
left a comment
daltino
left a comment
There was a problem hiding this comment.
This PR cleanly shifts multi-model frontend routing responsibilities from Gemini CLI to Antigravity CLI across several command files. The changes are consistent and follow the repo's conventions without introducing ambiguity. The adjustments align with the contribution guidelines, preserving workflow clarity and intelligent routing rules. Nice work!
2 participants
Summary
codeagent-wrapperalready ships a workingAntigravityBackend(verified live:--backend antigravityresolves toagy -p ...and returns a correct response), and~/.claude/.ccg/prompts/antigravity/*.mdrole prompts already exist inccg-workflow. Only these four command files still hardcoded--backend gemini.multi-frontend.md,multi-execute.md,multi-plan.md, andmulti-workflow.mdto route frontend calls through--backend antigravityinstead of--backend gemini, point role-prompt paths atprompts/antigravity/instead ofprompts/gemini/, and drop the gemini-only--gemini-modelflag (antigravity has no CLI equivalent;codeagent-wrapperpicks its own default model).multi-backend.mdis untouched (codex-only, no frontend routing).docs/*/commands/are intentionally left as-is (out of scope, maintained separately).Test plan
~/.claude/bin/codeagent-wrapper --backend antigravity "Reply with exactly the word: pong" .— resolved toagy -p ..., returnedpong, exit 0.grep -ri gemini commands/multi-frontend.md commands/multi-execute.md commands/multi-plan.md commands/multi-workflow.md— 0 matches.~/.claude/.ccg/config.tomlrouting instead of being hardcoded in the command markdown.