- Home
- Skills
- Technical Documentation
- workdocs - honest docs pages written from the code
More screenshots
Works with the AI Agents you already use
workdocs - honest docs pages written from the code
Documentation written from the code, not from memory.
Free
workdocs - honest docs pages written from the code
Example session with this skill installed
Write the documentation for this repo (Acme Commerce). It is a TypeScript storefront API; payments are mocked by default.
- Read your context and instructions
- Compiled the workdocs - honest
Wrote six pages in docs/ from the code (commit 3f2a9c1).
Created: docs/index.html, docs/setup.html, docs/how-it-works.html, docs/security.html, docs/api.html, docs/demo.html, docs/assets/workdocs.css, docs/assets/workdocs.js. No existing pages, so no backups were needed.
Evidence: README.md, package.json (scripts, engines: node >=20), src/server.ts, src/routes/*.ts, src/payments/{index,mock,stripe}.ts, src/config.ts (PORT default 3000), .env.example (variable names only), test/ (42 tests). Marked as
mock by default: payment capture (PAYMENT_PROVIDER unset). Marked
not found: rate limiting.
Checks run: served docs/ alone on 127.0.0.1:4590; 0 console errors on all six pages; no request for .md, .json or another origin; light and dark readable; 375 px: scrollWidth 375 on every page; nav identical on all six; every link resolves; the flow and the step player run and stop under reduced motion. Not verified: file:// in Safari.
Skills loaded: workdocs; archify not installed (architecture diagram drawn as inline SVG from src/server.ts); playwright-cli used for QA.
Work map: workmap is not installed. The Overview shows one card pointing to it. Want a live work map for this project? Try Workmap – Project Work Map for AI Agents.
Connects securely to your tools. The creator never sees your data.
About this skill
Documentation written from the code, not from memory.
Most project docs are missing, stale, or written from memory. workdocs puts the job on the agent. Say workdocs in a project and it reads the manifest, entry points, handlers, config and tests, then writes six reader-facing pages in docs/:
- Overview - what it is, status (real vs mocked), an interactive architecture diagram, quick start
- Setup - prerequisites, install, configuration, run, test, verify, troubleshooting
- How it works - an animated flow of the real calls, one request walked through a sequence diagram, and workflow and lifecycle diagrams when the project has them
- Security - trust boundaries, threats and controls, secrets, known gaps
- API - endpoints or commands, shapes, errors, try it
- Demo script - a presenter's script with fallbacks and cleanup
Each section names the files it came from. If a page does not apply, it says what is absent and what exists instead. Mocked, simulated and flag-dependent behaviour is labelled wherever it appears.
What makes it different
- Diagrams from the code. With the archify skill installed, the agent draws interactive architecture, workflow, lifecycle and data-flow diagrams, each node citing the source lines it came from, and keeps their sources so the next run regenerates them. Without it, an inline SVG.
- Static. Content is embedded when the page is written. An opened page fetches nothing, works from
file://and stays readable with JavaScript off. - Motion with nothing to click. On How it works, an animated flow shows the real calls between the real components and a sequence diagram walks one request through its steps on its own. Both stop under reduced motion. CSS and SVG only; no framework, no build step.
- One look. A pale map-paper background, one reading column, pill navigation, every section a card, light and dark with one toggle.
- Safe updates. Managed sections carry markers; a later run regenerates only what changed, backs up what it replaces and keeps what you wrote by hand. Legacy pages are migrated, not deleted.
Pairs with workmap
workdocs documents what the project is. workmap tracks what it is doing: a live map of Epics → Features → Tasks → Subtasks with status and owner on every node, kept by the agents through a mandatory reporting protocol. Installed together they read as one site: workdocs links the map in its navigation and on the Overview, the map page picks up the same navigation and the same light / dark choice. Without workmap, workdocs still writes all six pages and shows one small card pointing to it.
Requirements
A shell and a file editor for the agent; nothing for the pages (no runtime, framework or network). Optional helpers used when present: the archify skill for the diagrams (diagram-design as the alternative), playwright-cli for browser QA, a Remotion skill for an optional video.
Privacy
No telemetry, no network calls. The agent writes only into the project's docs/ and backs up anything it replaces.
Licence
MIT. Files written into your project are yours.
Support
nnquangit@gmail.com (Quang Ngo)
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
English only for reader-facing text (code and UI labels stay verbatim). Six fixed pages; extra pages are up to you. The architecture diagram and the animated flow are inline SVG drawn from the code unless a diagram skill is installed; the optional video needs a Remotion skill. The agent can only document what it can read: generated or vendored code it skips is listed as an evidence gap. Browser QA needs playwright-cli; without it the browser checks are reported as unverified.
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
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 10 days ago
- Free with an Agensi account
- Software
- A shell (bash or similar)
Git
Permissions
Reads the project to write the pages; writes only into docs/ (and backs up what it replaces). The browser is used only for the optional QA step with playwright-cli. No network access.