feat: Extends performance domain with anti-pattern scanning skills

[MajorLift]

Motivation

domains/performance is four extension pattern guides (perf-hooks-effects, perf-rendering, perf-state-management, perf-react-compiler) and one mobile advisory skill. The patterns are documented. What none of them does is fire without being sought.

A selector returning a fresh object reference each call produces no build error, no failing test and no user report. It produces a component that renders 178 times instead of 61. Nobody searches for a guide to a problem they do not know they have, which is what makes a review-time scan a different thing from a better-written guide.

The guides are also one platform each: the four perf-* skills install into metamask-extension only, the performance skill into metamask-mobile only. The same antipattern had two unlinked treatments and no shared definition.

Overview

Two review-time scans for the React and Redux antipatterns that cause render cascades, plus the shared definitions both scans cite. Supersedes #82 and #49. Profiling, render deltas and benchmark design are in #144.

• effect-antipattern-scan and selector-antipattern-scan: grep-driven checklists over a PR diff. Both carry metamask-extension and metamask-mobile overlays.
• knowledge/: effect-antipatterns, selector-antipatterns, render-cascade.
• The mobile performance skill's mm-* references map each pattern onto the codebase instead of redefining it. No mobile-specific content was removed.

Coverage gap this surfaced: no skill here detects renderItem identity instability, and no overlay mentions FlatList or FlashList.

Showcase

A render-cascade epic in metamask-extension, both arms measured with WDYR: 1d6d1811b8 178 re-renders → 0d37542a08 61. Its two constituent fixes, #39312 and #37147, are both selectors returning a fresh reference per call, which is the pattern these scans look for.

Trial runs against merged PRs nobody flagged, each posted beside its diff. #39310 "149 consumers" counts call sites, not renders · #40859 deep-equal to reference-equal, recomputations may rise · #44392 ?? {} fresh object, unmemoized, into useSelector · mobile#33578 per-render const feeding a FlashList.

The same epic counted by selector evaluations rather than renders: 49 → 2, and a single transition a timer cannot resolve, at #39314.

What I could not check: the per-slice reference-stability framing in the cascade file, against the store's actual behavior.

Sources: frontend-performance.md.

[MajorLift]

Context budget

What this PR costs an agent, measured from an install rather than read from the diff. Three tiers, and only the first is unavoidable.

Skill	Frontmatter	Selected	+ refs & knowledge
data-analysis	221 chars	~1,428 tok	~1,428 tok
effect-antipattern-scan	89 chars	~1,193 tok	~1,193 tok
extension-profiling	134 chars	~680 tok	~680 tok
react-render-delta	780 chars	~1,938 tok	~1,938 tok
selector-antipattern-scan	100 chars	~2,510 tok	~2,510 tok
benchmark-design	109 chars	~882 tok	~882 tok

Frontmatter is the only tier paid unconditionally — every agent loads it on every run once the skill is installed, used or not, because it is what the agent reads to decide relevance. The 28 skills across the eleven open skill PRs sit at a median of ~1,716 tokens selected and ~1,860 with references followed. All are within the 1,536-character description budget.

Selected is paid only when the agent picks the skill. + refs & knowledge is the ceiling if every bundled reference is then read; it is a worst case, not an expectation.

Method

tools/install --repo metamask-extension --maturity experimental against this branch at c18c751a7, measured per installed skill directory. Repo overlays are merged into the emitted SKILL.md, so they land in the selected tier rather than being missed by a source-byte count. Token figures are bytes/4 — a proxy for scale, not accounting.

These figures are pinned to the commit above and drift on every push; #96 tracks automating them.