which-tier

Routes code changes to the cheapest verification tier that can actually detect them.

39|7|Updated Jul 3, 2026
One-click install
npx skills add https://github.com/GLinnik21/plx-native --skill which-tier-glinnik21
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: which-tier
Source: https://github.com/GLinnik21/plx-native/tree/main/.agents/skills/which-tier
Command: npx skills add https://github.com/GLinnik21/plx-native --skill which-tier-glinnik21

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Verification suites pass even when they are structurally blind to the change being tested, producing false confidence. This Skill decides which verification tier — host tests, simulator, or physical television — can actually see a given change, and names the tiers that cannot. ## Core Features & Use Cases - Tier Routing: Maps what a change touched (pure logic, UI layout, text rasterization, animation, frame rate, player pipeline, FFI, release configuration, packaging) to the cheapest tier capable of verifying it. - Blind-Spot Analysis: Documents what each tier structurally cannot detect, such as the fps tier's blindness to pixels and the synthetic test tier's bypass of the Plex metadata-to-plan pipeline. - Handoff to Executors: Ends by delegating to execution skills like ui-sim, tv-session, tv-lock, fleet-plan, profile-tv, crash-triage, or cut-release. - Use Case: After editing an icon's geometry in a Rust TV app, ask which suite to run — the Skill routes you to a device capture looked at by eye, not the fps suite that cannot see pixels. ## Quick Start Ask which verification tier is needed for your change, describing what files or behavior you modified.

Frequently Asked Questions about which-tier

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I know which test suite to run for my change?

Route by what the change touched, not by convenience. Pure logic needs only host tests, UI layout needs simulator screenshots plus one device capture, and player pipeline changes need the synthetic or server-backed device suites.

Why did my tests pass but the bug still shipped?

A tier that cannot see your change passes rather than fails. For example, the fps tier only measures frame rate and is blind to pixel content, so visual bugs like incorrect watched-mark rendering pass all 11 fps scenes green.

Can the simulator verify frame rate or text rendering?

No. The simulator uses a different GPU, driver, and FreeType than the television, so it cannot answer frame rate or text rasterization questions. Simulator heartbeats carry a sim=1 flag so they are never mistaken for device measurements.

What does a bare ./tests/run.py run actually test?

Since 2026-08-22 a bare run executes the synthetic tier with generated clips and no Plex server. It bypasses the metadata-to-plan pipeline entirely and cannot see resume, markers, Up Next, track selection, or transcode paths.

When do I need the physical television for verification?

Use the device only when lower tiers are structurally incapable: text rasterization, frame rate, video playback, and anything only visible on the panel itself. Take the tv-lock first and wake the set before running any device suite.