🛠️ Systems Engineering Skills

100 free AI skills for systems and hardware engineers.

Review requirements against ECSS. Check electrical and mechanical interfaces. Build power and mass budgets. Review environmental qualification tests, FMEA and verification evidence.

Browse all 100 skills → · Setup · Example prompts

Arc Skills task bench

In Codex, install Arc Skills, then ask “Open Arc Skills.” Pick a task, choose files or paste evidence, and press Start in Codex. The existing Codex conversation runs the selected skill; the panel shows actual saved results and new reports. Try a sample supplies a synthetic requirements example. No Arc account, Program, hosted API or model API key is required.

The panel searches all 100 skills from this catalogue. Inputs are copied into the local toolkit workspace (8 MiB per picked file, at most six files). Task requests, results and new artifacts stay there across restarts. Codex and its AI provider still process content used in the conversation. Existing source files are not overwritten, and engineering results are proposals rather than approvals.

If the host cannot send a message, Copy prompt gives the exact handoff for the same conversation. Refresh results reads what the assistant actually saved; it never fabricates findings or marks an unfinished task complete. If a send times out, check the conversation before using Resume in Codex. When the host supports local file opening, saved reports open from the panel; otherwise their workspace-relative paths are shown.

The task bench is the skills-first direction. The separately packaged hosted Arc workspace below remains preserved and is not required or removed by this update.

For the live Arc product inside Codex, see the separate Arc workspace plugin. It adds authenticated hosted model tools and the existing Arc interface; these 100 skills and their local file toolkit remain independent.

🍽️ Pick a task

There are also skills for interfaces, architecture, budgets, safety, reliability, change control and design reviews. See the full menu →

🚀 Get started

For a local MCP client, install uv and choose a dedicated workspace folder. On macOS or Linux:

mkdir -p "$HOME/Documents/arc-engineering-work"
codex mcp add arc-engineering -- uvx --no-config --from arc-engineering-mcp==0.2.1 arc-engineering serve --workspace "$HOME/Documents/arc-engineering-work"

Other local MCP clients use uvx with the same arguments after --. The folder must exist. Start a fresh conversation and ask Open Arc Skills or request a specific engineering task. The following source/ZIP route is also available.

Download this repo using Code → Download ZIP. Open the unzipped folder in Codex, or attach the ZIP to a ChatGPT conversation that supports file analysis.

Copy and paste this setup prompt:

Set up Arc Skills from:
https://github.com/ArcHelps/systems-engineering-skills
Use the repository or ZIP I provided, or fetch the public repository if your tools allow it.
Read README.md and docs/SETUP.md, then choose the route this client supports.
If you can run local commands, install the locked dependencies, connect the local
MCP to Codex using the documented configuration, and check skill discovery.
Use a dedicated arc-engineering-work folder for inputs and new outputs.
Keep my existing settings and source files intact. Do not publish or upload my files.
If you cannot run a local MCP, read the attached catalogue and load only the
selected skill and its references for this conversation. Explain any unavailable tools.
Show me five relevant skills and run the synthetic requirements example if possible.
Say which route actually worked and whether I need to restart or enable anything.

Local MCP setup needs Python 3.11+ and uv. No model API key needed. PyPI installation and registry release status. Want only one task? Install an individual self-contained skill using your assistant’s available tools. This does not install the MCP.

💬 Try it

Attach a file or place it in your configured work folder, then paste a prompt:

Clean requirements

Use Arc Skills to clean and map customer-requirements.xlsx.
Preserve every source ID and original statement. Show the field mapping,
duplicate IDs, TBDs/TBCs, missing information and unresolved links.
Give me a cleaned workbook and an import proposal. Reconcile all row counts.

Review a specification

Use Arc Skills to extract and review this requirements PDF.
Keep page references, notes, figures and deleted requirements visible.
Give me a requirements table and the most consequential issues first:
contradictions, broken references, unclear limits and unverifiable statements.
Separate confirmed problems from ambiguities. Don't invent missing values.

Review against ECSS

Use Arc Skills to review REQ-014 against the ECSS-E-ST-10-06 edition
I attached. Cite the inspected criteria, explain each finding and propose
the smallest correction. Preserve the original wording beside the proposal.
If you cannot inspect the applicable standard, label the review advisory.

Build a power budget

Use Arc Skills to build a system power budget from component-loads.xlsx.
Separate standby, nominal and peak demand by operating mode. Preserve units,
quantities, duty cycles and source values. Show demand against the supplied
power limits, conversion losses, missing loads and remaining margin.
Don't assume every peak load operates simultaneously or invent missing values.

No file yet?

Use Arc Skills to review this requirement:
"The unit shall provide an appropriate warning quickly."
Explain what is missing and propose clearer wording without inventing a latency.

🔒 Your files

The local MCP processes files in your chosen folder without uploading them to an external server. Your assistant and its AI provider still process the content used in the conversation. ChatGPT uploads follow ChatGPT's policies. Data flow →

The skills and code are free under the MIT license. Your AI assistant may have its own charges. Standards are not bundled; provide authorized copies when needed. Engineering edits are proposals for your review, with source wording preserved.