Launch a Perpetua-Tools agent with LM Studio endpoint validation and env setup
npx skills add https://github.com/diazMelgarejo/Perpetua-Tools --skill agent-run이 명령을 Claude Code에 복사하여 붙여넣어 스킬을 설치하세요
| name | agent-run |
| description | Launch a Perpetua-Tools agent with LM Studio endpoint validation and env setup |
| disable-model-invocation | true |
1. Verify .env.lmstudio exists:
[[ -f .env.lmstudio ]] || python3 ~/.openclaw/scripts/discover.py --force
cat .env.lmstudio | grep LM_STUDIO
2. Validate Win endpoint:
source .env.lmstudio 2>/dev/null || true
curl -s --connect-timeout 3 "$LM_STUDIO_WIN_ENDPOINTS/v1/models" \
| python3 -c "import sys,json; d=json.load(sys.stdin); print(len(d['data']), 'models on Win')"
If curl fails → python3 ~/.openclaw/scripts/discover.py --force then retry once.
3. Launch:
set -a && source .env && source .env.lmstudio && set +a
python agent_launcher.py "$@"
Commandeer AlphaClaw/OpenClaw runtime defaults. Set environment profiles, backup/restore sessions, enumerate live agents, self-heal connectivity issues. Run when starting any OpenClaw-dependent session.
Startup scenario classification, probe retry, history-driven adaptive timeouts, cloud fallback, and .env.local override detection for Perpetua-Tools and orama-system. Use when startup fails to reach backends, when you need to understand which scenario the system is in (FULL_DISTRIBUTED, MAC_DUAL, MAC_OLLAMA_ONLY, MAC_LMS_ONLY, CLOUD_ONLY, FULLY_OFFLINE), or when routing behaves unexpectedly after reboot.
Development conventions and patterns for Perpetua-Tools. Python project with mixed commits.
Query live state across all 3 repos — versions, branch status, LM Studio endpoints, file manifests. Run whenever you need a situational snapshot of the stack.
Crystallize session learnings into LESSONS.md and instincts. Runs automatically at session end (Option C — proposes updates, commits only after user approval). Also invocable on-demand.
Development conventions and patterns for Perpetua-Tools. Python project with mixed commits.