Official descriptionOpen source7d ago
sectapunterx/deep-e2e-auditor
Claude Code skill: deep functional + visual (eyes) e2e audits of desktop and web apps
View on GitHub ↗Direct link · no redirect page
- Star count
- 1 ★ · sectapunterx/deep-e2e-auditor · read from the repository page · (checked 2026-10-16 20:00:20)
- Source
- read directly from its GitHub page · type: open-source project
- Harvest history
- first seen 2026-10-10 · last seen 2026-10-10 · seen 1× · listed by 1 sources
- Score
- score 103 · 1 ★ · 1 sources · repo
- Install
pip install pillow
FAQ (answers come from our records)
How do I install it?
The repository provides this install command:
pip install pillow. It comes from the repository itself; we cannot guarantee it runs as-is in your client.Who is it for?
We filed it under these tags: 开源项目. The tags come from the entry itself, not a taxonomy we imposed.
Who maintains it?
repository
sectapunterx/deep-e2e-auditor；first catalogued 2026-10-10；last seen 2026-10-10。There is no official certification or quality rating here — we record sources and dates only.Third-party quality signals
deep-e2e-auditor
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
A Claude Code skill that runs a deep end-to-end audit of a desktop app (Windows, macOS, Linux) or
a web app/site. You get a tiered report and an annotated screenshot gallery covering two things:
- **Functional:** every feature is exercised through the real UI down to storage, network and
integrations. Parallel area agents run in isolated sandboxes, and every serious finding is
re-checked independently.
- **Idiot test:** if the user can do it wrong, they will. Every on-screen hint is followed
literally in every focus state, every global key is pressed with every layer open, actions are
doubled, abandoned half-way and done out of order, and every state must have a keyboard way back.
- **Use it:** some bugs exist only for someone who is working. The product is used for real, on
content holding every kind of element: every element is clicked every way, every edge between
two elements is crossed (Backspace, Delete, selection, copy), every mode is switched with input
still pending, and the screen is checked again after autosave and debounce timers fire.
- **Eyes:** every surface is captured in every state, and same-kind surfaces are compared side by
side. This catches the context menu that doesn't match the others and the overloaded card. It
also includes a heuristic walkthrough, pixel measurements, and persona sessions that use the
product like a person would.
Install
On Windows the path is %USERPROFILE%\.claude\skills\deep-e2e-auditor.
Use
In Claude Code, inside the project you want audited:
The skill reads the repo first, then asks one round of questions: scope, sandboxes, what must
never be touched, and real-data permission. It then proves a harness, maps features and surfaces,
and fans out agents. It writes everything to .deep-e2e/runs/ / in the project, which is
git-excluded locally and never committed.
What's inside
| Path | |
|---|---|
| SKILL.md | the workflow |
| references/ | safety rules, intake, harness recipes per platform/framework, functional method, eyes method, report and verification |
| templates/ | agent briefs (functional, eyes surfaces, persona, verification), finding schema |
| scripts/win/ | capture (all windows of a process, occlusion-proof), input (posted or real), UI Automation tree |
| scripts/mac/, scripts/linux/ | the same four jobs for macOS (JXA + screencapture) and X11/Xvfb (xdotool + ImageMagick + AT-SPI) |
a web app/site. You get a tiered report and an annotated screenshot gallery covering two things:
- **Functional:** every feature is exercised through the real UI down to storage, network and
integrations. Parallel area agents run in isolated sandboxes, and every serious finding is
re-checked independently.
- **Idiot test:** if the user can do it wrong, they will. Every on-screen hint is followed
literally in every focus state, every global key is pressed with every layer open, actions are
doubled, abandoned half-way and done out of order, and every state must have a keyboard way back.
- **Use it:** some bugs exist only for someone who is working. The product is used for real, on
content holding every kind of element: every element is clicked every way, every edge between
two elements is crossed (Backspace, Delete, selection, copy), every mode is switched with input
still pending, and the screen is checked again after autosave and debounce timers fire.
- **Eyes:** every surface is captured in every state, and same-kind surfaces are compared side by
side. This catches the context menu that doesn't match the others and the overloaded card. It
also includes a heuristic walkthrough, pixel measurements, and persona sessions that use the
product like a person would.
Install
On Windows the path is %USERPROFILE%\.claude\skills\deep-e2e-auditor.
Use
In Claude Code, inside the project you want audited:
The skill reads the repo first, then asks one round of questions: scope, sandboxes, what must
never be touched, and real-data permission. It then proves a harness, maps features and surfaces,
and fans out agents. It writes everything to .deep-e2e/runs/ / in the project, which is
git-excluded locally and never committed.
What's inside
| Path | |
|---|---|
| SKILL.md | the workflow |
| references/ | safety rules, intake, harness recipes per platform/framework, functional method, eyes method, report and verification |
| templates/ | agent briefs (functional, eyes surfaces, persona, verification), finding schema |
| scripts/win/ | capture (all windows of a process, occlusion-proof), input (posted or real), UI Automation tree |
| scripts/mac/, scripts/linux/ | the same four jobs for macOS (JXA + screencapture) and X11/Xvfb (xdotool + ImageMagick + AT-SPI) |
- Source
- github · https://github.com/sectapunterx/deep-e2e-auditor
- Browse by field
- Open source
- Traction
- 1 ★
- added
- 7d ago
More in Open source
Filed under
Every tag opens the full list for that filter. Counts are the live catalogue figures.