mcporbital-openclaw-space — OpenClaw Skill
Chat-channel face of MCPOrbital. Drop the folder into your OpenClaw runner; users in Telegram, Discord, Slack-style topics get /space …slash commands routed into the MCP server, with an immutable per-workflow audit trail.
What ships
mcporbital-openclaw-space/
├── SKILL.yaml
├── README.md
├── routing/
│ ├── intent_to_tool.yaml
│ └── topic_policy.yaml
├── agents/
│ ├── planner.md
│ ├── researcher.md
│ ├── critic.md
│ └── reporter.md
└── templates/
├── brief.md.tpl
└── refusal.md.tplNo code — YAML routing + markdown role prompts + handlebars templates. Reviewable end-to-end, no opaque binaries.
Drop & restart
# On the OpenClaw VPS — drop the skill into the plug-in dir cd /root/.openclaw/workspace/plugins/ git clone https://github.com/lysechko1/mcporbital.com-openclaw-skill mcporbital-openclaw-space # Configure secrets (per OpenClaw conventions) vim mcporbital-openclaw-space/.env # MCPORBITAL_MCP_URL=https://mcp.mcporbital.com # MCPORBITAL_MCP_JWT=eyJ... # Restart the OpenClaw runner
The skill is fully self-contained — it does not modify any other OpenClaw plug-in or shared configuration. Secrets live in the standard .envpicked up by the runner.
User surface
| Command | What | Role gate |
|---|---|---|
| /space brief <object> [to #channel] | Research brief | any |
| /space card <object> | Compact object profile | any |
| /space risk-check <sat> [next N d] | Conjunction risk | Team+ |
| /space weather | Live snapshot | any |
| /space weather YYYY-MM-DD | Retro snapshot | any |
| /space papers <query> | Literature list | any |
| /space agent-team <task> | Multi-agent workflow | Team+ |
| /space subscribe kp >= N to #ch | KP threshold alert | Team+ |
| /space help | Command list | any |
Planner → Researcher → Critic → Reporter
Each role has bounded MCP-tool access. The OpenClaw runner enforces JWT scopes so a researcher cannot accidentally call a planner-only tool. Every step writes a structured event to audit_trail/<date>_<trace_id>.jsonl; daily SHA-256 manifest gives tamper-evidence.
Same as MCP server
Topic-level rate caps. Role-gated intents. Hard refusal patterns matched server-side. Audit trail immutable and tamper-evident.