skillmesh
Scannednpx machina-cli add skill varunreddy/SkillMesh/skillmesh --openclawHow to use
- Ask:
Route this request with SkillMesh: <your task> - Run the router script to emit routed context.
- Continue the task using only the returned top-K cards.
Command
Use scripts/route.sh (or scripts/route.py) to emit context:
scripts/route.sh --provider codex --registry <path> --query "<query>" --top-k 5
Role Commands
List available roles and install a role bundle (role card + missing dependency cards):
skillmesh roles
skillmesh roles list
skillmesh Data-Analyst install
skillmesh roles install Data-Analyst
Use --dry-run to preview install changes.
If skillmesh is not on PATH, use:
scripts/roles.sh (interactive)
Parameters
--registry: Registry file path (tools.yaml,tools.json, orroles.registry.yaml).--provider:codex(default) orclaude.--top-k: Number of cards to retrieve (default5).--backend:auto,memory, orchroma(defaultauto).--dense: Enable optional dense reranking.--instruction-chars: Max instruction chars per expert (default700).--query: User request text.
Notes
- If
--registryis omitted, the router usesSKILLMESH_REGISTRY. - If
skillmeshis not onPATH, the router falls back topython -m skill_registry_rag.
Source
git clone https://github.com/varunreddy/SkillMesh/blob/main/skills/skillmesh/SKILL.mdView on GitHub Overview
SkillMesh retrieves the most relevant SkillMesh cards from a registry and injects a Codex-ready context block before coding. It’s ideal for tasks that span multiple domains, enabling top-K routing instead of loading the full skill catalog to speed up and focus guidance.
How This Skill Works
You request routing by phrasing: Route this request with SkillMesh: <your task>. The router script (scripts/route.sh or scripts/route.py) emits a routed context containing the top-K cards. You then proceed using only the returned cards, keeping the workflow concise and targeted.
When to Use It
- A task spans multiple domains and requires cross-domain guidance.
- You want to avoid loading the full skill catalog by using top-K routing.
- You need a Codex-ready context block injected before coding.
- You have a known registry path for skills and want to fetch from it (via --registry).
- You want to preview/install skill changes with dry-run or manage roles before execution.
Quick Start
- Step 1: Route this request with SkillMesh: <your task>
- Step 2: Run the router script to emit routed context
- Step 3: Continue the task using only the returned top-K cards
Best Practices
- Start with the default top-K (5) to balance coverage and noise.
- Ensure the registry path is correct and accessible (tools.yaml, tools.json, or roles.registry.yaml).
- Use --provider codex (or default) to generate Codex-ready context.
- Use --dry-run to preview changes before applying installations.
- Validate that the returned top-K cards collectively cover all required domains for the task.
Example Use Cases
- Routing a complex data science task that touches SQL, Python, and visualization tools to get cohesive guidance.
- Selecting documentation and tooling for a software integration across multiple services.
- Coordinating a multi-domain DevOps deployment with relevant operational cards.
- Preparing a chatbot execution plan by routing top-K expert cards for the task.
- Injecting a Codex-ready context block prior to coding a robotics control sequence.
