sluglistUse cases
One contract, four ways in
Feedback moves from whoever found the problem to whoever fixes it. Both ends can be a person or an agent, and in every combination the thing that travels between them is the same: a folder of plain files with a documented format. Pick the end you are standing at.
Whoever finds it
reports
- A client signing off a release
- A PM or tester walking a checklist
- A customer hitting a bug in production
- You, on your own laptop
- A QA agent driving a real browser
The artifact contract
a folder of plain files
session-2026-08-16-a1b2/ session.yaml 01-save-does-nothing.md 01-save-does-nothing.png fixes.yaml
Versioned, additive-only, documented field by field.
Whoever fixes it
resolves
- A developer opening the report
- Anyone the single-file HTML gets forwarded to
- A fix agent that localizes and patches
- Any script that can read files
You, with the app running on localhost.
Click the bug on your running localhost app and it becomes a file: screenshot, selector, component, console errors, action trail — in a local folder your coding agent reads.
- Describing a bug costs one click
- Nothing leaves the machine
- Your agent stops guessing where
The people who decide whether it shipped correctly — the client, the PM, the tester.
Put a build on staging with a checklist of what shipped. The client, PM or tester walks it, checks items off, flags problems visually — and you get a coverage map, not a chat thread.
- A coverage map instead of “looks good 👍”
- Nothing for them to install or sign up for
- You do not write the list by hand
Real users in your beta or production app.
Let real users report problems with an annotated screenshot — inputs masked, PII scrubbed from text, delivery through an endpoint you own. MIT, self-hosted, no accounts.
- The report arrives with the context you would have asked for
- Privacy is the default, not a setting you remember
- It cannot become your problem
A repo where an agent can run the app in a browser and edit the code.
A QA agent walks the checklist in a real browser and writes evidence-backed verdicts; a fix agent answers them; a re-test round closes the loop — and sluglist status decides when to stop.
- Evidence, not claims
- The loop stops for the right reasons
- You get an audit trail, not a “done”