Skip to content

v1.0.0 — Three Skills, One Professional Stack

Latest
Latest

Choose a tag to compare

View all tags
@igor-holt igor-holt released this 28 May 09:25
· 34 commits to main since this release
v1.0.0
a912ea2

GitHub Release Notes Draft — v1.0.0 Bundle

Repository: https://github.com/igor-holt/openclaw-skills
Tag: v1.0.0
Title suggestion: v1.0.0 — Three Skills, One Professional Stack: grok-persistent-state + mcp-openclaw-bridge + smithery-mcp-orchestrator

Release Summary

This is the coordinated initial public release of three OpenClaw skills designed from the ground up to compose into a complete professional agent infrastructure layer on the OpenClaw + Smithery + MCP ecosystem.

  • grok-persistent-state — Persistent project state and unified omniskill registry for stateless LLMs.
  • mcp-openclaw-bridge — Authoritative patterns and templates for turning MCP servers into permanent ClawHub skills.
  • smithery-mcp-orchestrator — Production-grade Smithery discovery, connection management, and secure token scoping for autonomous agents.

These skills were extracted and hardened from real Genesis Conductor deployments, clawbot-grok reference implementations, and multi-agent orchestration work. They prioritize correctness, explicit safety boundaries, token scoping, durable memory, and composability over hype.

All skills are published individually on ClawHub under MIT-0 and are intended to be installed together.

ClawHub install commands:

clawhub install grok-persistent-state
clawhub install mcp-openclaw-bridge
clawhub install smithery-mcp-orchestrator

Full documentation, SKILL.md files, state templates, mapping examples, and combined workflows: https://github.com/igor-holt/openclaw-skills

What's Included

1. grok-persistent-state v1.0.0

  • Canonical project_state.json schema (goal, current_focus, recent_changes, open_tasks, decisions, artifacts, smithery metadata, etc.)
  • Optional human-readable PROJECT_STATE.md
  • Safe read_project_state + patch_state workflow with strict injection discipline
  • Unified tool/skill registry loading pattern
  • Zero network surface; everything remains strictly local
  • Explicit trigger phrases and MUST / MUST NEVER safety rules (never store secrets or PII)

2. mcp-openclaw-bridge v1.0.0

  • Repeatable thin-adapter pattern for publishing MCP capabilities as native ClawHub skills
  • mcp-mapping-example.json and recommended SKILL.md structure
  • Mandatory privacy & data-flow declarations ("data that stays local")
  • Support for stdio, HTTP, and SDK-based MCP servers
  • Designed to complement native OpenClaw skills and pair with the other two skills in the bundle

3. smithery-mcp-orchestrator v1.0.0

  • Namespace discipline and persistent tracking of active namespaces
  • Full MCP connection lifecycle handling (connected / auth_required / error)
  • Advanced token scoping examples using policies and rpcReqMatch restrictions
  • Tool discovery (search / list / get) and JSONL piping patterns
  • Explicit guidance for human handoff on interactive auth flows
  • Direct composition recipes with grok-persistent-state (memory) and mcp-openclaw-bridge (promotion)

How the Three Skills Compose

The bundle implements a closed professional loop:

Discovery & Security (smithery-mcp-orchestrator)
→ Memory & Continuity (grok-persistent-state)
→ Distribution & Permanence (mcp-openclaw-bridge)

Example production workflow (detailed in docs/combined-workflow.md and examples/combined-three-skill-workflow.md):

  1. Use smithery-mcp-orchestrator to create a dedicated namespace and scoped connections (GitHub, Linear, etc.) with narrow token policies.
  2. Immediately record namespace, connection IDs, and token metadata via grok-persistent-state.
  3. On subsequent turns, restore exact context automatically — no re-authentication or re-explanation required.
  4. Identify high-value repeated MCP tool combinations and promote them into standalone ClawHub skills using the mcp-openclaw-bridge patterns.

This turns ephemeral agent experimentation into reliable, auditable, reusable infrastructure.

Installation & Verification

After publishing to ClawHub (see publishing/RECOMMENDED_COMMANDS.md and publishing/publish-all.sh):

clawhub install grok-persistent-state
clawhub install mcp-openclaw-bridge
clawhub install smithery-mcp-orchestrator

Inspect published skills:

clawhub inspect grok-persistent-state --versions --files
clawhub search "persistent state"

See individual skill READMEs and SKILL.md files for trigger tables, full workflows, and safety declarations.

Philosophy & Design Goals

  • Thermodynamic verification + topological orchestration mindset applied to practical agent tooling (Genesis Conductor lineage).
  • Reliability and verifiability over convenience.
  • Explicit security boundaries, especially around credentials and token scoping.
  • Everything that can be made permanent and distributable (via ClawHub) should be.
  • Strong preference for narrow, short-lived, auditable permissions.
  • Full transparency: every safety rule, mapping, and composition pattern is documented in the SKILL.md files.

Attribution & Licensing

Maintained by Igor Holt
AI Architect | Creator of Genesis Conductor
https://github.com/igor-holt/openclaw-skills

All skills and supporting files are released under the MIT-0 License (public domain equivalent).

Special thanks to the OpenClaw and Smithery communities for the substrate that makes these patterns possible and valuable.

Next Steps & Feedback

  • Report production usage, edge cases, and suggested improvements via GitHub Issues.
  • Future minor releases will maintain backward compatibility with the v1.0.0 trio.
  • Contributions that improve safety declarations, add new verified workflow examples, or extend the state schema while preserving the strict boundaries are welcome.

Full repository: https://github.com/igor-holt/openclaw-skills

End of GitHub Release draft. Copy the content above (excluding this line) into the GitHub Releases UI when creating the v1.0.0 tag.

Assets 2
Loading