← Back to Skills Marketplace
Provenote MCP Outcome Workflows
by
Yifeng[Terry] Yu
· GitHub ↗
· v1.1.1
· MIT-0
111
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install provenote-mcp-outcome-workflows
Description
Teach an agent to install Provenote's first-party MCP server, connect it in a host, and run read-first outcome workflows.
Usage Guidance
This skill appears coherent for teaching a local Provenote MCP workflow, but before installing you should: (1) verify the provenance of the GitHub repo it asks you to clone (https://github.com/xiaojiou176-open/provenote.git) and review its code, (2) ensure the 'uv' tool and 'git' are available and trustworthy on your host, (3) run the server in an isolated/sandbox environment if possible (it will execute code from the cloned repo), and (4) update the provided OPENCLAW/OPENHANDS config snippets to point only at safe, intended paths. The metadata omission about required binaries (uv/git) is minor but worth correcting; if you cannot review the external repo, avoid launching the server on sensitive infrastructure.
Capability Analysis
Type: OpenClaw Skill
Name: provenote-mcp-outcome-workflows
Version: 1.1.1
The skill bundle instructs the agent to clone and execute an external MCP server from a third-party GitHub repository (xiaojiou176-open/provenote) using shell commands (bash -lc) and 'uv run', as detailed in references/INSTALL.md and references/OPENCLAW_MCP_CONFIG.json. While these capabilities are plausibly required for the stated purpose of managing research workflows, the reliance on unverified remote code execution via shell scripts constitutes a high-risk behavior. No explicit evidence of intentional malice, such as data exfiltration or backdoors, was found within the provided files.
Capability Assessment
Purpose & Capability
The name/description match the instructions: the packet teaches how to install, launch, and wire a first‑party Provenote MCP server and run read‑first workflows. One inconsistency: INSTALL.md lists 'uv' as a required binary and a specific GitHub repo to clone, but the registry metadata reported 'required binaries: none' — the skill will in practice need 'git' and 'uv' (and network access) to follow the install steps.
Instruction Scope
SKILL.md and the references clearly confine actions to cloning the Provenote repo, launching its MCP server (via 'uv run provenote-mcp'), editing host MCP config snippets, and performing read-first tool calls (draft.list, research_thread.list, auditable_run.list). The instructions do not ask the agent to read unrelated system files or exfiltrate credentials, but they do require the agent/host to run shell commands and fetch code from an external GitHub repo (https://github.com/xiaojiou176-open/provenote.git).
Install Mechanism
This is instruction-only (no install spec), which lowers install-surface risk. However, the INSTALL.md guides cloning an external GitHub repository and running 'uv sync' / 'uv run', which will pull and execute code from that repo — a normal pattern for installing software but one that requires the host reviewer to vet the external source and the 'uv' tool.
Credentials
The skill declares no required environment variables or credentials and its runtime instructions do not reference secrets. It does require filesystem and shell access on the host (editing/pointing configs at a local clone and launching the server), which is appropriate for the stated purpose.
Persistence & Privilege
No elevated platform privileges are requested: always:false, user-invocable, and the packet does not attempt to modify other skills or system-wide agent configs. It only instructs configuring the host's MCP server entries (which is expected).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install provenote-mcp-outcome-workflows - After installation, invoke the skill by name or use
/provenote-mcp-outcome-workflows - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.1
- Added a TROUBLESHOOTING guide for improved user support.
- Updated documentation, including README and reference materials, for clarity and usability.
- Minor adjustments to configuration and supporting files to reflect the latest usage patterns and best practices.
v1.1.0
Version 1.1.0
- Expanded installation and usage guidance with new reference docs, including INSTALL.md, DEMO.md, CAPABILITIES.md, and sample host configs.
- Updated skill description and triggers to clarify scope and supported workflows.
- Restructured README and SKILL.md for clearer agent guidance on installing, wiring, and using Provenote MCP in host environments.
- Enhanced step-by-step workflow instructions and added a suggested first prompt for users.
- Formalized success criteria and boundaries for outcome workflows.
v1.0.0
Initial release of provenote-mcp-outcome-workflows.
- Provides host-facing access to Provenote's first-party MCP outcome workflows.
- Enables safe, read-first exploration before allowing limited write actions.
- Supports operations like listing drafts, research threads, and auditable runs.
- Offers controlled mutation endpoints: research_thread.to_draft, draft.verify, draft.download, auditable_run.create, auditable_run.download.
- Intended for public-ready use but not yet listed as a live marketplace or catalog skill.
Metadata
Frequently Asked Questions
What is Provenote MCP Outcome Workflows?
Teach an agent to install Provenote's first-party MCP server, connect it in a host, and run read-first outcome workflows. It is an AI Agent Skill for Claude Code / OpenClaw, with 111 downloads so far.
How do I install Provenote MCP Outcome Workflows?
Run "/install provenote-mcp-outcome-workflows" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Provenote MCP Outcome Workflows free?
Yes, Provenote MCP Outcome Workflows is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Provenote MCP Outcome Workflows support?
Provenote MCP Outcome Workflows is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Provenote MCP Outcome Workflows?
It is built and maintained by Yifeng[Terry] Yu (@xiaojiou176); the current version is v1.1.1.
More Skills