- Home
- Skills
- Technical Documentation
- Technical Handoff Writer: C4 + Simple English
More screenshots
Works with the AI Agents you already use
Technical Handoff Writer: C4 + Simple English
Your AI agent writes a handoff or design doc with C4 diagrams in simple English, and checks it before you share it.
$15
Technical Handoff Writer: C4 + Simple English
Example session with this skill installed
Write a handoff document for this service for the team that will run it. Save it as handoff.html in this folder. Run the skill's checks before you finish.
- Read your context and instructions
- Compiled the technical handoff writer
- Generated the UI component
This is a sample service: a worker that makes photo thumbnails, 44 lines of Python. The agent saw it for the first time. Below is its real report, unedited. The screenshot in the gallery shows part of the document it wrote.
I wrote the handoff and saved it as handoff.html in this folder. I haven't published it or committed it. Both of the skill's checks pass.
Final check output
doc check PASSED — 3 figure(s), 1 table(s), 4 callout(s), 104 sentence(s)
exit=0
figure 1 flowchart ok -> /tmp/figs/handoff-1.png
figure 2 flowchart ok -> /tmp/figs/handoff-2.png
figure 3 sequenceDiagram ok -> /tmp/figs/handoff-3.png
3/3 rendered
exit=0
The first run also passed, but with 3 passive-voice warnings, which I rewrote. "Rendered" only means the diagram drew, so I also looked at the images and made two fixes:
- Figure 2: the text
DELETE_ORIGINAL = Truewas cut off in its box. - Figure 3: the caption pointed to the wrong step.
The PNGs in /tmp/figs were only for checking and aren't part of the handoff.
What the handoff tells the team, all taken from the code:
It can't be undone: DELETE_ORIGINAL = True deletes each original photo after making thumbnails, and no other copy exists. The handoff gives the steps to follow before changing sizes or the resize code.
Messages that can never succeed: the worker deletes the original before it removes the queue message. If it crashes between those two steps, every retry fails because the original is gone. Such messages, and some in uploads-dead, must not be put back on the queue.
One error stops everything: handle() doesn't catch errors, so one corrupt photo stops the whole worker. Nothing in the repo restarts it, and there are no metrics or alerts. The handoff proposes three alerts.
The 60-second timeout is a guess: the worker takes 10 messages at a time, and each one comes back after 60 seconds whether or not it's finished. Nobody has measured how long a photo takes, so the handoff says not to run two workers until someone does.
Gaps it flags: the storage and imaging modules aren't in this repo, so the worker can't run from it alone. Also, the handoff is based only on the 44 lines of code; nothing here has been run or tested.
handoff.html
TSX · React component
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
A handoff fails in two ways. The diagrams are hard to read. Or the text is smooth, and the reader never learns what can break.
What you get
Your AI agent reads your code and writes one HTML document. It contains:
- A status note at the top. It lists what has not run yet, so nobody reads the design as a working system.
- C4 diagrams in Mermaid: the system and its neighbors, the parts inside it, and one scenario step by step. Green is yours, grey is external, and tan is a person, on every diagram.
- A states table. Each state has a meaning and a recovery step.
- A callout for the one action that nobody can undo, with the exact steps to do it safely.
- Text in Simplified Technical English (ASD-STE100): short sentences, active voice, and one name for each thing.
Two checks before you share it
check_doc.pystops on text left from the template, a missing status note or callout, an empty recovery step, wrong figure numbers, and sentences that are too long. It warns about passive voice and filler words.check_diagrams.pydraws every diagram and stops on a syntax error.
Works in any agent
In Claude, the agent publishes the document as an Artifact. In other agents, you get an HTML file. It opens in any browser and draws its own diagrams.
Included
- The skill instructions.
- A styled template.
- A complete example document for a fictional service.
- The two check scripts.
What you need
- Python 3.10 or later.
- Node.js 18 or later for the diagram check. The first check downloads mermaid-cli. This takes a minute or two.
- Internet access to show the diagrams. The page loads Mermaid from a CDN.
The diagram check needs Node.js and a headless browser, and some sandboxes block it. The diagrams need internet access to show. The agent writes only what it finds in your code and notes, and it marks the rest as not verified. The writing rules are for English text.
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 any payment.
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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 28 days ago
Requires
Permissions
Terminal: the agent runs the two check scripts. Read files: the agent reads your code and the template. Write files: the agent writes one HTML file, and the diagram check can save PNG images to a folder you choose. Network: the diagram check downloads mermaid-cli from npm on its first run, and the finished page loads Mermaid from cdn.jsdelivr.net.