| name | check-rico-ai |
| description | Read-only audit of Rico Hunt AI provider routing. Verify RICO_AI_PROVIDER env var, fallback chain, and provider health without calling live AI APIs. |
| triggers | ["user","model"] |
| allowed-tools | ["read","grep","exec"] |
check-rico-ai
Read-only audit of Rico Hunt AI provider routing. This skill never calls live OpenAI, DeepSeek, HuggingFace, or JSearch APIs. It only inspects code and environment variables.
What it verifies
RICO_AI_PROVIDER is set to a supported provider (openai, deepseek, huggingface, or keyword fallback).
- The active provider matches the fallback chain described in
CLAUDE.md:
deepseek → HuggingFace → keyword/templated fallback
openai → fallback
src/rico_openai_agent.py routing logic reflects the provider chain.
src/rico_env.py health report fields (*_key_present, ready_for_*, hf_available) are computed at runtime, not static env vars.
- Required API keys are present as env vars but their values are not exposed.
Quick checks
echo "RICO_AI_PROVIDER=$RICO_AI_PROVIDER"
echo "OPENAI_API_KEY set: $([ -n "$OPENAI_API_KEY" ] && echo yes || echo no)"
echo "DEEPSEEK_API_KEY set: $([ -n "$DEEPSEEK_API_KEY" ] && echo yes || echo no)"
echo "HF_TOKEN set: $([ -n "$HF_TOKEN" ] && echo yes || echo no)"
grep -n "RICO_AI_PROVIDER\|provider\|ready_for_\|hf_available" src/rico_openai_agent.py src/rico_env.py
Files to read
src/rico_openai_agent.py — provider routing logic
src/rico_env.py — health report dataclass and runtime checks
CLAUDE.md — AI Provider Routing section (source of truth)
Safety constraints
- Do not print actual API key values.
- Do not call
openai.chat.completions.create, DeepSeek, or HuggingFace endpoints.
- Do not set
RICO_AI_PROVIDER to a different provider without approval.