SystemWright — Design Reusable AI Workflows
Turn vague goals into reusable AI workflows with clear context, tool boundaries, and review loops.
Free
Works with the AI Agents you already use
SystemWright — Design Reusable AI Workflows
Example session with this skill installed
I want AI to help me manage my company, but I do not know what to build. Start with a weekly decision review using facts I provide manually. Keep decisions and external actions under my control.
- Read your context and instructions
- Compiled the systemwright
Illustrative design walkthrough, not a production transcript or a measured benchmark.
Goal
Turn a broad request to manage a company with AI into one weekly decision review. Begin with manually supplied facts; no company integrations are required for the first version.
Inputs
Bring this week's metrics, customer feedback, blocked work, last week's commitments, and the current business priority. Missing evidence remains marked as missing.
Work system
| Layer | First version |
|---|---|
| Prompt | Act as a decision-support assistant. Separate facts, interpretations, and decisions. |
| Context | Use the supplied weekly facts and prior decisions. Cite each material claim to its input. |
| Harness | Work inside the conversation. Do not send messages, assign owners, or change company systems. |
| Loop | Collect facts, draft the brief, get the founder's decisions, and check outcomes next week. |
Weekly deliverable
Produce a concise factual summary, the most consequential risks, one to three decisions with alternatives, proposed next steps, and questions that need a human answer. Do not invent metrics, owners, or deadlines.
Human decisions
The founder chooses priorities, approves responsibilities and external commitments, and decides which proposals to reject. Unclear evidence is escalated instead of treated as certainty.
Trial and review
Run one review using a real week of facts. Pass the review only if every factual claim has a source, the decisions are concrete, and next week's checks are explicit. If the process does not help the founder make useful decisions, revise the inputs or narrow the scope before adding tools.
Record for next week
Save the supplied facts, accepted decisions and reasons, open questions, approved next steps, and review findings. After several useful manual runs, consider integrations only for demonstrated repetitive work.
Connects securely to your tools. The creator never sees your data.
About this skill
SystemWright helps you design a reusable way to work with AI. Start with a messy goal or a process you repeat, and work toward a clear brief, a small first version, and checks you can use to judge the results.
It organizes the design around four layers:
- Prompt: the task, intended outcome, and decision rules.
- Context: the facts, examples, and records the AI needs.
- Harness: tools, permissions, and verification.
- Loop: repetition, feedback, budgets, and stop conditions.
Use it to plan a recurring research routine, support-triage process, founder review, or team workflow. It can help decide whether a prompt is enough or whether a skill, tool integration, or automation is justified.
You receive a plain-language brief, a work-system design, a copyable starting prompt, and a way to trial and review the workflow. The skill designs the system; external integrations and live automations still need to be implemented and verified separately.
Includes SKILL.md, seven reference guides, host metadata, and the MIT license. The source is public and free. Designed for Agent Skills-compatible hosts; platform-specific delivery and access to bundled reference files need to be supported by the host.
Source and documentation: https://github.com/EthanQC/SystemWright License: MIT. Free and open source.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Design guidance, not a hosted automation service. Proposed tools and integrations need separate implementation and testing. Output quality depends on the supplied context and host model. Requires access to bundled reference files. External actions remain subject to the user's authorization and the host's permissions.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 24 days ago
- Free with an Agensi account
Permissions
Read access is for bundled references and any context files you choose to supply. Writing a Markdown plan or record is optional and only needed when you ask the agent to save one. No shell, network, browser, environment-variable access, or account connection is required for the core design conversation.