Agent skill
multi-ai
Use when configuring ARC_PROVIDER_MODE or routing verification work to external AI providers (Codex, Gemini). Claude-only is always the safe fallback.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/multi-ai-cmtkdot-arc
SKILL.md
Multi AI
Lightweight provider routing for Arc. Controls when and how external AI providers (Codex, optionally Gemini) participate alongside Claude in the workflow. No command may fail solely because optional providers are unavailable — Claude-only is always the safe fallback.
Invocation Contract
Inputs:
ARC_PROVIDER_MODEARC_MULTI_AI- Provider availability/auth state
Outputs:
- Effective provider mode
- Explicit downgrade notice when fallback is applied
Provider Modes
Set via .arc/config.json (preferred) or environment variables (override):
claude-only(default) — All work stays with Claude. No external calls.codex-assist— Codex produces draft findings; Claude validates and publishes.codex-primary— Codex performs primary analysis; Claude runs final adjudication.
Resolution order: ARC_PROVIDER_MODE env var > .arc/config.json provider_mode field > default (claude-only).
The multi_ai: true flag (in config) or ARC_MULTI_AI=true (env) must be set to enable any non-Claude mode. If provider mode is set to a non-Claude mode but multi_ai is false or unset, emit an explicit downgrade notice and continue in Claude-only mode.
Provider Detection
Detection runs in this order before any external call:
- Check
ARC_MULTI_AI=true(env) ormulti_ai: true(.arc/config.json) — if false/unset, use Claude-only. - Detect
codexCLI availability — is the binary on PATH? - Detect Codex authentication —
OPENAI_API_KEYor active Codex auth session. - Detect optional Gemini — only if the current command explicitly supports it.
If any required check fails, downgrade to Claude-only and continue. Log the downgrade reason.
Agent Routing Matrix
Claude-Only Agents (Never Routed Externally)
plan-creator-defaultopenspec-defaultbeads-defaultbead-workerfix-finding- Team lead orchestration
Reason: these agents mutate project artifacts and must remain deterministic. External provider variability could corrupt plan/spec/bead state.
Codex-Eligible Agents
verify-typescriptverify-performanceverify-architectureverify-securityverify-requirements
In codex-assist mode:
- Codex produces draft findings.
- Claude validates and normalizes findings against Arc's verification schema.
- Claude publishes the final result.
In codex-primary mode:
- Codex performs primary analysis with full context.
- Claude runs a final adjudication gate before reporting results.
External Call Rules
- All external provider calls must be explicit and logged in output.
- Recommended Codex invocation pattern:
codex exec --model ${ARC_CODEX_MODEL:-gpt-5.3-codex} "<prompt>" - On provider failure or timeout, fallback to Claude-only and continue.
- Never silently swallow provider errors — log them and note the fallback.
Gemini Integration
Gemini support is optional and command-specific. When available:
- Use for design review and frontend analysis (large context window advantage).
- Detection follows the same pattern: check availability, check auth, fallback if missing.
- Gemini is never required for any core Arc workflow.
Related Skills
routerdetermines the tier; provider routing applies within any tier.debatemay route debate participants to different providers in multi-AI mode.- Verification agents are the primary consumers of external provider routing.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
Didn't find tool you were looking for?