vscode-peacock
AGENTS.md1.2k starsChanged 2 months agoMITView on GitHub
Peacock is a Visual Studio Code extension that subtly changes the color of your workspace. It's ideal when you have multiple VS Code instances, use VS Live Share, or use VS Code's Remote features and want to quickly identify which editor is which. To run the extension locally, press F5 in VS Code — this launches the Extension Development Host with the extension loaded.
- Reads credentials
- Deletes or force-pushes
- Installs packages
Trending
More talk lately.about the whole repository
Who's talking about johnpapa/vscode-peacock
- microsoft/vscode-remote-release“Remote: Iteration Plan for July 2019”GitHub · Jul 5, 2019 · 7 reactions
- A Hacker News story“Visual Studio Code: Peacock”Hacker News · May 3, 2021 · 4 points
- forem/forem“Embed GitHub Code”GitHub · Mar 12, 2019 · 4 reactions
- @hejchristian.com“Desperately needed to install this Peacock extension today, as I find myself with three different projects …”Bluesky · Mar 21, 2026 · 2 likes
- zed-industries/zed“Allow per-project (worktree) theme overrides for Peacock-style window tinting”GitHub · May 20, 2026 · 1 reactions
The biggest 5 of 18 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 18
Talked about on GitHub, Hacker News, Bluesky and Stack Overflow (18 times).
On GitHub · 15
- Jul 5, 2019Remote: Iteration Plan for July 2019 — microsoft/vscode-remote-release7 reactions
- Mar 12, 2019Embed GitHub Code — forem/forem4 reactions · 16 comments
- May 20, 2026Allow per-project (worktree) theme overrides for Peacock-style window tinting — zed-industries/zed1 reaction
- Oct 31, 2019Improve Onboarding — Kruemelkatze/vscode-dashboard1 reaction · 6 comments
- Oct 23, 2026Multiple BOBs when more than 1 window is open — EricKhmel/BobPet0 reactions
10 more on GitHub
- Sep 17, 2026Modern UI doesn't work with Peacock extension — posit-dev/positron0 reactions · 3 comments
- Aug 4, 2026ModernUI breaks custom theming — bojordan/ColorIdentity0 reactions
- Mar 21, 2026feat: Kunterbunt – automated workspace color extension for VS Code — martinus/vsode-extension-kunterbunt0 reactions
- Sep 13, 2023Add Code of conduct — thoroc/projen-typescript-git-hooks0 reactions
- Apr 19, 2023devcontainer creation fails when project is empty — microsoft/vscode-remote-release0 reactions · 2 comments
- Sep 22, 2020Peacock — coder/code-server0 reactions · 2 comments
- Nov 29, 2019Reach out to extensions to adopt new extension kind behaviour — microsoft/vscode0 reactions · 1 comment
- Nov 2, 2019https://github.com/johnpapa/vscode-peacock/commits — johnpapa/express-to-functions0 reactions · 2 comments
- Apr 30, 2019Integrate GitHub dependabot — SAP/spartacus0 reactions · 3 comments
- Mar 15, 2019Change Title Bar Color on Pair Swap — FordLabs/pairodoro0 reactions · 1 comment
On Hacker News · 1
- May 3, 2021Visual Studio Code: Peacock4 points
On Bluesky · 1
- Mar 21, 2026Desperately needed to install this Peacock extension today, as I find myself with three different projects open and no way to distinguish… — @hejchristian.com2 likes and reposts · 2 replies
On Stack Overflow · 1
- Dec 7, 2020VSCode different Theme setting for each project in a WorkSpace1 vote
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository2 other files it gives its agents.
- .github/copilot-instructions.mdCopilot instructions
- ai-readySkill
The file
Show its rules and text
We split this AGENTS.md into the rules it gives an agent
17 rules. About 0 are common, 6 are distinctive, and 11 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
11
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 90
When migrating a host test to the unit lane, only move it if the underlying logic is genuinely pure — extract a pure helper first if needed.
- Distinctive: line 94
Every bug fix must include a regression test that fails without the fix and passes with it
- Distinctive: line 95
Every new feature must include unit tests covering the happy path and relevant edge cases
- Distinctive: line 96
Never merge code that reduces the passing test count
- Very specific: line 118
Add the entry under the matching category heading (### Features, ### Fixes, ### Docs, ### Tests, ### Infrastructure — create the heading if …
- Distinctive: line 120
Prefer adding the changelog entry as part of the PR itself before merging.
- Very specific: line 165
Add menu visibility rules in contributes.menus.commandPalette if needed
- Very specific: line 166
Add the command ID to the Commands enum in src/models/enums.ts
- Very specific: line 169
Add a unit test in src/test/
- Very specific: line 170
Update docs in docs/guide/ if the command is user-facing
- Very specific: line 178
Add the setting name to StandardSettings enum in src/models/enums.ts
- Very specific: line 179
Add a reader in src/configuration/read-configuration.ts
- Very specific: line 180
Add unit tests in src/test/
- Very specific: line 181
Update docs in docs/guide/
- Very specific: line 187
Add to AffectedSettings enum in src/models/enums.ts — and add the VS Code color token to ColorSettings enum
- Very specific: line 188
Add the property to IPeacockAffectedElementSettings in src/models/interfaces.ts
- Distinctive: line 216
Touching or adding a webview?: Run it against the Webview Security Checklist above — a real injection bug shipped in this repo's one …
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Peacock — Agent Guide
Project Overview
Peacock is a Visual Studio Code extension that subtly changes the color of your workspace. It's ideal when you have multiple VS Code instances, use VS Live Share, or use VS Code's Remote features and want to quickly identify which editor is which.
- Publisher: johnpapa
- VS Code Marketplace: Peacock
- Version: See
package.jsonfor the current version
Repository Structure
vscode-peacock/
├── src/ — Extension source code (TypeScript)
│ ├── extension.ts — Extension entry point (activation, command registration)
│ ├── commands.ts — Command implementations (enter color, random, favorites, etc.)
│ ├── apply-color.ts — Core color application logic
│ ├── color-library.ts — Color manipulation utilities (uses tinycolor2)
│ ├── configuration/ — VS Code configuration read/write helpers
│ ├── models/ — TypeScript interfaces, enums, constants, state
│ ├── inputs.ts — User input prompts (color picker, quick pick)
│ ├── statusbar.ts — Status bar color display
│ ├── live-share/ — VS Live Share integration
│ ├── remote/ — Remote development integration
│ ├── test/ — Mocha unit tests
│ ├── logging.ts — Output channel logging
│ ├── mementos.ts — Global state persistence
│ ├── notification.ts — User notifications
│ └── object-library.ts — Object/element management
├── e2e/ — Playwright end-to-end tests (docs screenshots)
├── docs/ — Docsify documentation site
│ ├── guide/ — User guide pages
│ ├── about/ — About pages
│ ├── changelog/ — Changelog page
│ ├── _sidebar.md — Navigation sidebar
│ └── index.html — Docsify entry point
├── resources/ — Extension icons and assets
├── testworkspace/ — Test workspace used by Mocha tests
├── .github/
│ ├── workflows/docs.yml — Docs deploy + Playwright e2e tests
│ ├── ISSUE_TEMPLATE/ — Bug report and feature request templates
│ └── PULL_REQUEST_TEMPLATE/ — PR template
├── webpack.config.js — Webpack bundler config (Node + Web targets)
├── tsconfig.json — TypeScript configuration
├── playwright.config.ts — Playwright e2e test config
└── package.json — Extension manifest, commands, settings, scriptsTech Stack
- Language: TypeScript (strict mode)
- Runtime: VS Code Extension Host (Node.js + Browser via
extension-web.js) - Bundler: Webpack (dual target:
extension-node.jsandextension-web.js) - Color library: tinycolor2 — all color manipulation goes through this
- VS Live Share:
vslspackage for Live Share integration - Linting: ESLint + Prettier (husky pre-commit hook runs Prettier automatically)
- Testing: dual-lane — Vitest (fast, pure-logic unit tests in
src/test/unit/) + Mocha (real VS Code extension host tests insrc/test/suite/) + Playwright (e2e docs tests ine2e/) - Docs: Docsify (static site in
docs/, deployed to GitHub Pages)
Build & Run
npm install # Install dependencies
npm run webpack # Build (development mode)
npm run vscode:prepublish # Build (production mode — webpack --mode production)To run the extension locally, press F5 in VS Code — this launches the Extension Development Host with the extension loaded.
Testing
npm test # Compile + run Mocha host tests + Vitest unit tests
npm run test:unit # Vitest only — fast, pure logic, no VS Code needed
npm run test:host # Compile + Mocha host tests + Live Share host tests
npm run just-test # Mocha host tests only (skip compile)
npm run test:e2e # Run Playwright e2e tests (docs screenshots)
npm run test:coverage # Run both lanes with coverage (host + Vitest)
npm run test-all # Run test + Live Share tests
npm run package:check # Package the VSIX and verify contents/size (see Release Process)Test structure — two lanes, pick based on whether the code touches the vscode API:
- Unit lane (
src/test/unit/, Vitest): pure logic with novscodeAPI dependency — color math, data transforms, pick-A-or-B decisions. Runs in ~1s, no VS Code needed.vscodeitself is mocked (src/test/unit/mocks/vscode.ts). - Host lane (
src/test/suite/, Mocha + Sinon): anything that touches the realvscodeAPI — commands, real config reads/writes, UI prompts. Boots an actual VS Code extension host; slower, but it's the ground truth. - E2e tests live in
e2e/and use Playwright to capture docs screenshots - The
testworkspace/directory is used as a VS Code workspace during host tests - When migrating a host test to the unit lane, only move it if the underlying logic is genuinely pure — extract a pure helper first if needed. Never delete host coverage without an equivalent unit replacement that exercises the same branches.
Test requirements:
- Every bug fix must include a regression test that fails without the fix and passes with it
- Every new feature must include unit tests covering the happy path and relevant edge cases
- Never merge code that reduces the passing test count
- UI/theme-affecting changes (status bar, title bar, activity bar, color tokens) should be verified on Windows, VS Live Share, and Cursor before considering the fix complete — automated tests alone haven't caught real-world issues here in the past
- Treat "this can't be tested" claims in a PR as a hypothesis to verify, not a fact to accept. Before agreeing,
grepsrc/test/for the exactvscodeAPI/state the new code depends on (e.g.env.remoteName, a config key, a command). If any existing suite already stubs/mocks that exact dependency (host lane cansinon.stub(vscode.env, ...); unit lane only has what's insrc/test/unit/mocks/vscode.ts), the claim usually means "untestable in the one lane I checked," not "untestable everywhere" — write the test in the lane that already supports it instead of skipping coverage. See PR #757 for a worked example (claim was wrong; host-lane precedent existed inremote.test.ts).
Key Patterns and Conventions
- Commands are registered in
extension.tsand implemented incommands.ts - Color application flows through
apply-color.ts→color-library.ts(tinycolor2) - Configuration is read/written via helpers in
src/configuration/ - State is managed through VS Code's
workspaceStateandglobalStateAPIs (seemementos.ts) - Models define all TypeScript interfaces and enums in
src/models/ - Dual output — the extension compiles to both Node (
extension-node.js) and Web (extension-web.js) via Webpack
CI/CD
- CI (
ci.yml): Triggered onpull_requestandpushtomain(ignores docs/markdown). Runs, in order: Lint → Build (test-compile) → Test (Host) (xvfb-run npm run just-test) → Test (Unit) (npm run test:unit) → Package contents check (npm run package:check). - Docs + E2E (
docs.yml): Triggered on push tomainand PRs whendocs/,e2e/,playwright.config.ts, or the workflow itself changes. Runs Playwright e2e tests, then deploysdocs/to GitHub Pages.
Merging Pull Requests
Every PR merge updates docs/changelog/README.md — no exceptions. This applies to every PR, not just feature/fix PRs: Dependabot bumps, chores, CI/infra tweaks, and docs-only changes all get an entry under the ## Unreleased section, not just user-facing changes. An empty ## Unreleased section at release time is a signal something was merged without a changelog entry.
- Add the entry under the matching category heading (
### Features,### Fixes,### Docs,### Tests,### Infrastructure— create the heading if it doesn't exist yet under## Unreleased). Dependency bumps (Dependabot or manual) go under### Infrastructure. - Link the PR and/or issue the entry corresponds to.
- Prefer adding the changelog entry as part of the PR itself before merging. If a PR was already merged without one (e.g. an upstream-authored Dependabot PR you merged as-is), add a follow-up commit to
mainimmediately after merging — don't defer it to release time. - This feeds directly into the Release Process below, which only finalizes/renames the
## Unreleasedsection — it should never be the first time an entry is written.
Release Process
This checklist is mandatory for every release, no exceptions — run through it in full even when the release feels small. Almost all of it is agent-executable — Claude (or any agent working this repo) should just do it, not ask a human to. Exactly one step is human-only: it's called out explicitly below, and it's the only one that requires stopping and waiting for a person.
Agent-executable — do these yourself, don't wait to be asked
CI, on every push to main (confirm green, don't assume):
- Lint —
npm run lint - Build —
npm run test-compile - Test (Host) — Mocha extension-host tests (
xvfb-run npm run just-test) - Test (Unit) — Vitest (
npm run test:unit) - Package contents check —
npm run package:check(catches dev-only file leakage into the VSIX and package-size regressions; seescripts/check-vsix-contents.js)
Check the latest push-triggered run of ci.yml on main — not just the last PR's run, since PR runs can be stale relative to what actually merged. If you want to double-check locally instead of trusting CI, rm -rf node_modules && npm ci then run the same commands above (host tests still need a real VS Code/Electron runtime, so they can only be verified via CI in most dev environments).
Then, in order:
- Check for blocking open work — any in-flight PRs or issues the release should wait for (e.g., an in-progress dependency-modernization chain). Don't release out from under unfinished work.
- Finalize
docs/changelog/README.md— rename the## Unreleasedsection to## X.Y.Z (YYYY-MM-DD), add a fresh empty## Unreleasedheading above it for whatever comes next. Every entry should link the issue/PR it closes, and separate entries by category (Features / Fixes / Docs / Infrastructure). Pick the version bump per semver: a new user-facing feature → minor; fixes/infra only → patch; a removed/renamed command, setting, or default → major. Use the actual date the tag will be pushed, not the date drafting started — if the release PR sits open for a day or two (e.g., waiting on other work to land first), update the date before merging, not after. - Bump
package.json'sversion— usenpm version X.Y.Z --no-git-tag-version(updatespackage.jsonandpackage-lock.jsontogether, no git tag yet). Version lives solely inpackage.json— there's no other file to sync at this step. - Open a PR with the changelog + version bump, verify CI is green, merge it to
main. - Before tagging, do one final full read-through of everything landing in this release — the entire changelog section top to bottom, plus any README/docs changes — checked against what actually merged this cycle. Do this as a single comprehensive pass, not by fixing issues one at a time only as the maintainer happens to spot them; if a mid-cycle PR added an entry after the changelog was first drafted (e.g., a fix that landed while the release PR was still open), confirm it got folded in before tagging, not left as a dangling
## Unreleasedentry. - Create and push the git tag for the release (e.g.
vX.Y.Z) once the merge commit is onmainand green. - Create the GitHub Release from that tag, using the changelog section as the release notes.
Update— no longer needed. The README uses a live shields.io badge (README.md's "Latest published version" linehttps://img.shields.io/visual-studio-marketplace/v/johnpapa.vscode-peacock) that reads the current version directly from the Marketplace API, so it never goes stale and there's nothing to sync after publishing. General principle: never hardcode a value in README/docs that changes every release (version numbers, dates, counts) — prefer a live badge or a generic phrasing that stays true across releases.- Draft the release's social media announcements (LinkedIn, X/Twitter, etc.) if the release warrants one — pull the headline changes from the finalized changelog. Drafting is agent-executable; posting is not (see below). Follow these conventions when drafting:
- Always include a 🦚 emoji — this is Peacock's repo, it should feel like Peacock.
- Always give a copy-paste-ready version, not just the text shown in chat. Save each platform's draft to its own file (e.g.
/tmp/peacock-x-paste-safe.txt,/tmp/peacock-linkedin-paste-safe.txt) so the maintainer can copy straight from a file instead of the chat UI, which can mangle emoji/formatting on copy. - LinkedIn strips Markdown on paste (
**bold**, backticks, etc. render as literal stray characters) — write LinkedIn drafts as plain text, using emoji and line breaks for emphasis instead of Markdown syntax. - Truly-empty blank lines between paragraphs often get collapsed on paste (observed on LinkedIn). Insert an invisible zero-width space (U+200B) on each blank line so it survives as a real line and paragraph breaks are preserved.
- Verify every factual/attribution claim against the actual source before drafting — don't imply the maintainer personally discovered, reported, or fixed an upstream bug unless that's literally true. Check the linked upstream issue/PR yourself (e.g. via
gh issue view/gh pr viewon the external repo) rather than inferring credit from the changelog wording alone; changelog phrasing written by an agent earlier in the release can itself be imprecise about who did what. - If the announcement includes before/after or comparison screenshots, verify both images represent the exact same UI state before publishing anything. Diff them at the pixel level (e.g. Python/PIL
getpixel/crop comparison) rather than eyeballing — incidental UI state at capture time (mouse hovering over a sash/divider, a focused vs. unfocused window, an open dropdown) can render extra colors/borders that have nothing to do with the feature being demonstrated, and look like a real bug if left in. If found, patch the incidental artifact out (or recapture) so the comparison isolates only the actual difference being illustrated.
Human-only — steps that need a person
- Run
vsce publish(or manually upload the packaged VSIX) to push the new version to the Marketplace. This needs the publisher's Marketplace personal access token, which only the maintainer holds — an agent shouldn't have it or use it autonomously. It's also one of two genuinely irreversible, externally-visible actions in the whole process — everything before it can be redone or reverted; this can't. - Review and post the social media announcements. An agent can draft them, but publishing public content under the maintainer's voice/identity is the maintainer's call — review for accuracy (don't let a drafted "hook" overstate or misrepresent a feature) and post it themselves.
Adding a New Command
- Define the command in
package.jsonundercontributes.commandswith acommandID andtitle(category: "Peacock") - Add menu visibility rules in
contributes.menus.commandPaletteif needed - Add the command ID to the
Commandsenum insrc/models/enums.ts - Implement the command handler in
src/commands.ts - Register the command in
src/extension.tsusingvscode.commands.registerCommand - Add a unit test in
src/test/ - Update docs in
docs/guide/if the command is user-facing - Update
docs/changelog/README.md
Adding a New Setting
StandardSettings (scalar values, not color tokens)
- Define the setting in
package.jsonundercontributes.configuration.propertieswith type, default, and description - Add the setting name to
StandardSettingsenum insrc/models/enums.ts - Add a reader in
src/configuration/read-configuration.ts - Add unit tests in
src/test/ - Update docs in
docs/guide/ - Update
docs/changelog/README.md
AffectedSettings (toggle whether a VS Code color token is colored)
- Define the setting in
package.jsonundercontributes.configuration.properties(boolean, default false) - Add to
AffectedSettingsenum insrc/models/enums.ts— and add the VS Code color token toColorSettingsenum - Add the property to
IPeacockAffectedElementSettingsinsrc/models/interfaces.ts - Wire into
getAffectedElements()insrc/configuration/read-configuration.ts - Wire into the appropriate
collect*Settings()function insrc/configuration/read-configuration.ts - Wire into
updateAffectedElements()insrc/configuration/update-configuration.ts - Add unit tests in
src/test/suite/affected-elements.test.ts - Update docs in
docs/guide/ - Update
docs/changelog/README.md
Webview Security Checklist
Peacock has one webview today (color-picker-html.ts/color-picker-webview.ts, the custom color picker panel). Any PR that adds or touches a webview must be checked against this list -- a real injection bug shipped in an earlier revision of that panel (fixed, but only after a maintainer-requested review caught it) because none of this was checked at the time:
- Never interpolate a caller-supplied string into HTML using only
isValidColorInput()/tinycolor-style "is this parseable" validation as the guard. tinycolor'srgb()/hsl()parsing is not anchored to the whole string, so a value likergb(1,2,3)" autofocus onfocus="..."passes validity checks while still breaking out of an HTML attribute. Before interpolating any value into a webview's HTML, normalize it to a known-safe shape first (e.g.getColorHex()'s tinycolor-formatted#rrggbb[aa], which can only ever contain#and hex digits) -- don't just gate on "is this a valid color," gate on "is this exactly the narrow shape I'm about to embed." - Don't leave a webview's CSP on
script-src 'unsafe-inline'if you can avoid it. Generate a per-render nonce (plainMath.random()-based, likegetNonce()incolor-picker-html.ts-- no need for Node'scryptomodule, which theextension-web.jsweb/webworker bundle can't resolve) and usescript-src 'nonce-${nonce}'plus a matching<script nonce="${nonce}">, so a future change that accidentally reintroduces an interpolated value can't also inject a second inline script. - A Quick Pick's
onDidSelectItemcallback fires on every highlighted item, not just the final selection. If a Quick Pick is extended with a non-data sentinel item (e.g. this repo'scustomColorPickerLabel, appended to the favorites list to open a picker panel), the highlight-preview callback must explicitly skip that sentinel -- parsing it as real data and feeding the result toapplyColor()will treat it as invalid input, which unapplies every current color the instant the item is merely arrowed over, before it's ever selected. - A webview panel can be closed three ways: Cancel, Apply, or the tab's own close button (
onDidDispose). All three must revert/resolve through the same code path (ideally the same serialized message queue used for regular messages -- seeserializeMessageHandler()incolor-picker-html.ts), not a separate direct call. RoutingonDidDispose's revert around the queue lets it race a still-in-flight message (e.g. a rapid color-well drag) and either leave a stale value applied or run before/after the wrong thing. Even with that, once the outcome is decided (settled), every other handler that mutates external state (e.g. a live-preview callback callingapplyColor()) must check that flag too and no-op if it's already true -- otherwise a preview message that happens to be queued behind the revert can still overwrite it, since "enqueued after" only guarantees run-order, not that an earlier handler bothered to check whether its result still matters. - Guard every
panel.webview.postMessage()call with a disposed check, not just the first one. A message can still be mid-flight (queued) when the panel closes; posting to an already-disposed webview throws, and sincepostMessageisn't usually awaited, that surfaces as an unhandled promise rejection rather than a caught error. If you touchsrc/test/suite/lib/fake-webview-panel.ts, keep itspostMessagethrowing after dispose (mirroring the real API) -- a fake that always succeeds can hide exactly this bug. retainContextWhenHiddenshould betruefor any webview with meaningful in-panel state (typed input, a live preview,acquireVsCodeApi()state). The default discards and later rebuilds the webview's DOM/script when the panel is hidden and re-shown (e.g. the user switches tabs and comes back), silently resetting anything not persisted -- for a small panel, the memory cost of keeping it alive is worth avoiding that.- Before marking any PR that adds or touches a webview as ready to merge, do one dedicated adversarial pass over it -- separate from the feature-specific testing that prompted the change. Treat every value that reaches the webview's HTML or script as attacker/workspace-controlled, not just "a color the user picked," and walk the panel's full lifecycle (hide/show, all 3 ways to close it, an in-flight async message racing any of those) rather than only the happy path a specific ask was testing. Iterative, feature-driven work (implement X, fix Y) does not surface this class of bug on its own -- it takes a deliberate, separate pass with "find problems" as the explicit goal. (This checklist itself was written only after a maintainer-requested review found 9 issues, including a script-injection bug, that several prior feature-focused passes over this same panel had all missed.)
Common Pitfalls
- Don't forget both output targets — the extension ships as both Node and Browser bundles. Test both if changing core logic.
- Color manipulation goes through tinycolor2 — don't use raw hex string manipulation; use the color-library helpers.
- Husky pre-commit runs Prettier — your code will be auto-formatted on commit. Don't fight it.
- Tests need a workspace — Mocha tests use
testworkspace/as the VS Code workspace. Make sure it exists. - Version is in package.json only — unlike multi-file version bumps in some projects, the extension version lives solely in
package.json. - The root
CHANGELOG.mdis a pointer file — the real changelog lives atdocs/changelog/README.md. Always edit the docs version, not the root file. - Touching or adding a webview? Run it against the Webview Security Checklist above — a real injection bug shipped in this repo's one webview before that checklist existed.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

