Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 17 additions & 0 deletions AGENTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,23 @@ This project should support Monroe's public story as a software engineer who bui
- Keep permissions minimal and user-triggered.
- Preserve local-history compatibility when changing saved item shapes.

## Agent skills

### Issue tracker

Issues and specs are tracked in StoneHub/webDevFeedbackExt GitHub Issues.
See `docs/agents/issue-tracker.md`.

### Triage labels

The repo uses the five canonical Matt triage labels.
See `docs/agents/triage-labels.md`.

### Domain docs

This is a single-context repository.
See `docs/agents/domain.md`.

## Verification

- Run `npm test` and `npm run check` before claiming behavior is ready.
Expand Down
10 changes: 10 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# Changelog

## 1.8.0 (Unreleased)

- Refocused the active product on Element capture, Region/PDF capture, History, and one explicit Agent Handoff.
- Removed the active Visual Edit and Add Content creators while keeping their historical Capture Records readable.
- Added canonical Capture Record constructors shared by Element and Region saves.
- Renamed the History JSON handoff to `Send to Codex` and gave each Downloads export a unique filename.
- Added bounded MCP inbox discovery and newest-valid import under the user's Downloads directory.
- Removed stale implementation plans and added current issue-tracker, triage-label, and domain-document instructions.
- Kept Feedback Sessions on the checkpointed experiment branch rather than shipping them in this release.

## 1.7.2 (Submitted August 3, 2026)

- Added Add Content mode for anchoring proposed text, image placeholders, lists, and safe HTML/embed-frame placeholders to existing page elements.
Expand Down
10 changes: 10 additions & 0 deletions CONTEXT.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
# Dev Feedback Capture Context

## Glossary

- **Page Capture.** Feedback captured from a browser-visible web page or PDF.
- **Element Capture.** A Page Capture focused on one identifiable page element and its surrounding evidence.
- **Region Capture.** A Page Capture focused on a selected visible region, annotations, and supporting context.
- **Capture Record.** One saved, portable feedback item with its request, evidence, and source context.
- **History.** The user-facing collection of saved Capture Records.
- **Agent Handoff.** An explicit local transfer of Capture Records to a coding agent for implementation and separate verification.
48 changes: 19 additions & 29 deletions QUICKSTART.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

Get Dev Feedback Capture running in a few minutes.

Chrome Web Store v1.7 is public. This source checkout prepares a behavior-neutral v1.7.1 discovery refresh; the latest GitHub Release ZIP remains v1.2.0 as a manual fallback.
The active product is the browser capture core: Element, Region/PDF, History, and one explicit Agent Handoff. Store and GitHub release notes in this repository preserve earlier submission evidence.

## 1. Install the extension

Expand Down Expand Up @@ -31,27 +31,7 @@ Optional for local PDFs:
5. Click a page element, add your note, and save it.
6. The in-page capture list starts expanded. Use **⌄** to collapse it; the compact list stays on the nearest viewport edge as you drag it, and **⌃** expands it again.

## 3. Preview a visual edit

1. Open an injectable webpage and choose `Visual` in the extension popup.
2. Click `Start Visual Edit`, then select one page element.
3. Drag the selected outline to move the element, or drag its corner handle to resize it.
4. Use Undo, Redo, or Reset as needed.
5. Add the implementation note and optional acceptance checks, then save the spec.
6. The live page is restored; the original/proposed evidence and requested mutations remain in local History.

## 4. Propose new page content

1. Open an injectable webpage and choose `Add` in the extension popup.
2. Click the existing element that should anchor the new block.
3. Choose Text, Image placeholder, List, or HTML/embed frame and select its placement.
4. Add filler content or explain what the new block should communicate or support.
5. Add optional acceptance checks, then save the insert spec.
6. The temporary block is removed from the live page while its anchor, intent, structured insert mutation, and before/proposed evidence remain in History.

The HTML/embed frame is a safe placeholder; it does not execute supplied HTML or load remote content.

## 5. Compile an annotated region spec
## 3. Compile an annotated region spec

1. Open the target page or PDF in the browser.
2. Open the extension popup and switch to `Region`.
Expand All @@ -61,30 +41,40 @@ The HTML/embed frame is a safe placeholder; it does not execute supplied HTML or
6. Describe the requested change and optionally add one acceptance check per line.
7. Save the visual change spec.

## 6. Export saved feedback
## 4. Export saved feedback

Open the extension popup and select `Open History & Export`. This extension-owned page works for captures from normal pages, PDFs, and other surfaces where the in-page panel is unavailable. From History, you can:

- Download one `AI Bundle` ZIP with `prompt.md`, structured feedback and page context, before/annotated PNGs, and `report.html`
- Download `JSON for MCP` for full payloads including crop image data
- Choose `Send to Codex` to place the selected handoff payload in the configured local Downloads inbox for MCP import
- Download a self-contained `HTML Report` with embedded region images
- Copy `Markdown` for issue trackers or docs
- Copy `AI Prompt` for ready-to-paste implementation instructions based on saved text and source context

AI Prompt is text-only. Use AI Bundle when the implementation handoff needs its numbered evidence images.

## 7. Give a local agent project-scoped feedback
## 5. Give a local agent project-scoped feedback

1. In History, choose `Download JSON for MCP`.
2. Configure the MCP companion with the absolute target project path and the folder containing that export.
3. Ask the agent to call `dev_feedback_import` with the exact JSON path. If the export contains multiple site/file groups, also provide the exact `storageKey` shown by the first rejected import.
1. In History, choose `Send to Codex` to download the current History handoff.
2. Configure the MCP companion with the absolute target project path and the browser Downloads folder.
3. Ask the agent to call `dev_feedback_import_latest`. If the handoff contains multiple site/file groups, provide the exact `storageKey` shown by the first rejected import.
4. The agent can call `dev_feedback_list`, `dev_feedback_get`, and `dev_feedback_build_brief`, implement changes with its normal project tools, then record progress with `dev_feedback_status_update`.

Codex setup is one command per target project:

```sh
codex mcp add dev-feedback -- node /absolute/path/to/webDevFeedbackExt/mcp/cli.mjs \
--project /absolute/path/to/project \
--inbox /absolute/path/to/Downloads
```

The extension places the file in the inbox; users do not need to move it manually. Implementation and verification remain separate steps.

Setup and security boundaries are in `docs/mcp-local-agent.md`.

## Need Help?

- Full docs: see `README.md`
- PDF capture issues on local files: check `Allow access to file URLs`
- Element mode unavailable: use `Region` mode on non-injectable browser surfaces
- Element capture unavailable: use `Region` mode on non-injectable browser surfaces
- Region capture saves viewport-only crops in v1, not full-page screenshots
Loading
Loading