skills/fleet-check/SKILL.md
Enforce free LLM fleet usage for all code generation
npx skillsauth add rudi193-cmd/aionic-claude-skills fleet-checkInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
ALL code generation MUST use the free LLM fleet. NEVER generate code directly using Claude tokens.
Do NOT write code directly. Pause and verify fleet usage.
Option A - llm_router (Preferred):
import sys
sys.path.insert(0, 'C:/Users/Sean/Documents/GitHub/Willow/core')
import llm_router
llm_router.load_keys_from_json()
response = llm_router.ask('''
[Code generation prompt]
''', preferred_tier='free')
if response:
# Add proof-of-fleet comment
code = f'# Generated by: {response.provider} (llm_router)\n{response.content}'
# NOW write it
Option B - Kart Agent:
POST http://localhost:8420/api/agents/chat/kart
{"message": "Create [description]"}
ALL fleet-generated code MUST include proof comment:
# Generated by: cerebras (llm_router)
// Generated by: gemini-2.5-flash (llm_router)
Before writing, confirm:
If user asks you to generate code directly:
❌ BLOCKED: Cannot generate code with paid tokens
REQUIRED: Use free fleet instead
- Cost: Fleet = $0.00, Claude = real money
- Providers: Cerebras, Gemini, Groq, OCI, Ollama (14 total)
Would you like me to:
1. Generate via llm_router? (I'll show you the command)
2. Delegate to Kart agent?
3. Switch to Claude Code CLI? (Has enforcement hooks)
If attempting to write code without fleet:
STOP - Fleet usage required
I cannot write code directly. This would burn expensive Claude tokens.
Let me use the free fleet:
[Show llm_router command]
Every time you use fleet instead of direct generation:
NONE. Always use fleet for code generation.
This skill automatically activates when:
development
Use this skill when writing new features, fixing bugs, or refactoring code. Enforces test-driven development with 80%+ coverage including unit, integration, and E2E tests.
tools
Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction.
tools
System status — atoms, edges, ratio, boot health, unpushed work, open threads
tools
Manual session startup — when boot hooks are degraded, run the bridge-open sequence manually