Description

Per ADR #57. Removes committed agent skills/rules now provided by the centralized Consensys/skills repo. Adds yarn skills to install them on demand.

• All committed agent skills/commands/rules deleted from .cursor/, .claude/, and .agents/. Synced sets land via yarn skills; nothing under those dirs is tracked anymore.
• .gitignore blocks the agent dirs entirely. Only .cursor/BUGBOT.md, .cursor/hooks.json, and .cursor/worktrees.json are carved out (IDE/bugbot config that must stay tracked).
• yarn skills calls scripts/skills-sync.sh, which requires CONSENSYS_SKILLS_DIR to be set; falls back to setup instructions if not configured.
• .skills.local (per-engineer SKILLS_DOMAINS= config) is gitignored; .skills.local.example is committed as a template.

Kept (in-repo, not under agent dirs)

• docs/ trees (perps, predict, readme), AGENTS.md files
• docs/testing/{unit-testing,e2e-testing,component-view-tests}.md (see next section)

In-repo testing guides (docs/testing/)

.cursor/BUGBOT.md is read by Cursor's BUGBOT service, which cannot run yarn skills to fetch the centralized mms-* set. Several AGENTS.md files also mark the testing guides ALWAYS-mandatory and need a stable in-repo path.

To keep BUGBOT enforcement working without re-importing the deleted skill structure, the three test guides live as plain docs under docs/testing/:

• docs/testing/unit-testing.md
• docs/testing/e2e-testing.md
• docs/testing/component-view-tests.md (consolidated SKILL.md + 3 references)

The same guidance also reaches IDE harnesses via Consensys/skills + yarn skills. These docs are the canonical fallback for tools that can't run the sync.

Engineer setup (one time)

git clone git@github.com:Consensys/skills.git ~/dev/Consensys/skills
export CONSENSYS_SKILLS_DIR=~/dev/Consensys/skills   # add to shell rc

Then in this repo:

yarn skills                              # interactive prompt
SKILLS_DOMAINS=perps,testing yarn skills # non-interactive

Companion PRs

• Consensys/skills#2 — source of truth, install CLI, sync wrapper.
• MetaMask/metamask-extension#42488 — parallel migration.
• MetaMask/decisions#162 — ADR amendment.

Out of scope

• Postinstall auto-trigger.
• TS rewrite of CLI.
• Personal-skills tooling.

Changelog

CHANGELOG entry: null

Related issues

Fixes:

Manual testing steps

Feature: yarn skills sync from Consensys/skills

  Scenario: engineer installs skills with CONSENSYS_SKILLS_DIR set
    Given the engineer has cloned Consensys/skills locally
    And CONSENSYS_SKILLS_DIR points to that clone
    When the engineer runs `SKILLS_DOMAINS=perps yarn skills`
    Then 2 perps skills are installed under `.claude/skills/`
    And `git status --short` reports a clean working tree
    And `cat .claude/skills/mms-fix-perps-bug/SKILL.md` shows the managed banner

  Scenario: engineer runs yarn skills without CONSENSYS_SKILLS_DIR
    Given CONSENSYS_SKILLS_DIR is unset
    When the engineer runs `yarn skills`
    Then the script prints setup instructions and exits non-zero
    And no files are written under `.claude/`, `.cursor/`, or `.agents/`

  Scenario: tracked Cursor config carve-outs remain
    Given the repo is freshly checked out
    Then `.cursor/BUGBOT.md`, `.cursor/hooks.json`, and `.cursor/worktrees.json` are tracked
    And no other files under `.cursor/`, `.claude/`, or `.agents/` are tracked

Screenshots/Recordings

No UI change — agent tooling and repo wiring only.

Before

N/A

After

N/A

Pre-merge author checklist

• I've followed MetaMask Contributor Docs and MetaMask Mobile Coding Standards.
• I've completed the PR template to the best of my ability
• I've included tests if applicable
• I've documented my code using JSDoc format if applicable
• I've applied the right labels on the PR (see labeling guidelines). Not required for external contributors.

Performance checks (if applicable)

• I've tested on Android
  • Ideally on a mid-range device; emulator is acceptable
• I've tested with a power user scenario
  • Use these power-user SRPs to import wallets with many accounts and tokens
• I've instrumented key operations with Sentry traces for production performance metrics
  • See trace() for usage and addToken for an example

For performance guidelines and tooling, see the Performance Guide.

Pre-merge reviewer checklist

• I've manually tested the PR (e.g. pull and build branch, run the app, test code being changed).
• I confirm that this PR addresses all acceptance criteria described in the ticket it closes and includes the necessary testing evidence such as recordings and or screenshots.