speed-tiers - SKILL.md Agent Skill

name: speed-tiers description: "Time-based evaluation. What registers in 10 seconds, 2 minutes, 20 minutes? Read-only diagnostic." user_invocable: true

Speed Tiers — Atomic Skill

Purpose

Evaluates the site at three speeds. What registers in 10 seconds (a glance at the homepage)? What lands in 2 minutes (a browse through one tier)? What rewards 20 minutes (reading the full Governance trilogy)? Most visitors give you 10 seconds. A few give you 2 minutes. Almost nobody gives you 20. The site needs to work at all three speeds — and pull visitors from one tier into the next.

Read-only diagnostic. No file changes.

Usage

/speed-tiers              # Evaluate all 3 time tiers

Context

Read

Homepage (index.md or systemworks layout output) — above the fold is the 10-second test
_data/index.json — manifesto and route structure
1 project page per tier (first listed artifact in each route) — the 2-minute test
Full Governance trilogy (Savepoint, Aetherwright, Formwork) OR full Output section — the 20-minute test
Screenshots from .audit/screenshots/ if available (the 10-second test is best done visually)
DESIGN-SYSTEM.md if it exists — visual intent context

Invoke (as needed)

None. Reads pages directly.

Lens Criteria (embedded)

The three speed tiers below ARE this skill's criteria.

Evaluate

Tier 1: 10 Seconds (the glance)

Evaluate the homepage above the fold. A stranger has arrived. They will decide in 10 seconds whether to stay.

Identity: Do they know who this person is? (Name, what they do — even vaguely)
Hook: Is there something that makes them want to scroll or click? (A striking phrase, an interesting visual, an unexpected element)
Differentiation: Does this look like "another portfolio" or does it feel distinct? (See /the-walk — what room does the homepage content build? See /the-walls — what room does the ambient frame build? At 10 seconds, the frame registers before the content.)
Navigation clarity: Can they tell where to go next?
Visual first impression: Does the design itself communicate something? (Craft, care, personality — or generic template)

Tier 2: 2 Minutes (the browse)

Evaluate the homepage plus one click-through (a project page or tier index).

Manifesto landing: Does the homepage manifesto communicate the practice clearly enough to earn the click?
Project page delivery: Does the first project page they visit deliver on the homepage's promise?
Voice consistency: Same person speaking on both pages?
Depth signal: After 2 minutes, does the visitor sense there's more worth exploring? Or did they get the gist and that's enough?
Range signal: Do they understand this person works across domains? Or do they think it's a single-discipline portfolio?

Tier 3: 20 Minutes (the deep read)

Evaluate the full Governance trilogy or the full Output section as a sustained reading experience.

Vocabulary payoff: Does the material vocabulary (holds, breaks, drifts, scaffold, fidelity) make more sense by the end? Or is it still opaque?
Intellectual reward: Does the depth feel rewarding? Does the reader think "this person really thinks about this" or "this is over-engineered"?
Through-line: Does the sustained reading reveal a coherent practice, or disconnected projects?
The Aetherwright payoff: For someone who reads the whole Governance section — does the symbolic system (sigils, altitude, the Order) feel earned by the end? Or still alienating?
Return signal: After 20 minutes, does the visitor want to come back? Bookmark? Share?

Transition Hooks

The most important thing this skill evaluates: what pulls a visitor from one tier to the next?

10s → 2min: What on the homepage makes someone click deeper? Is it the manifesto? A project name? A visual element?
2min → 20min: What on the first project page makes someone want to read more? Is there a trail of breadcrumbs?
If either transition hook is missing, the site is effectively only the tier above it.

Output

Print in conversation. No file changes. Format:

# Speed Tiers — [COMPELLING / ADEQUATE / BRITTLE]

## 10 Seconds — [COMPELLING / ADEQUATE / BRITTLE]
**What registers:** [what a stranger absorbs in a glance]
**What's invisible:** [what they'd never know from 10 seconds]
**Hook to stay:** [what pulls them into 2 minutes — or nothing]

## 2 Minutes — [COMPELLING / ADEQUATE / BRITTLE]
**What lands:** [what a browser understands after homepage + one click]
**What's invisible:** [what requires more time]
**Hook to go deeper:** [what pulls them into 20 minutes — or nothing]

## 20 Minutes — [COMPELLING / ADEQUATE / BRITTLE]
**What rewards:** [what the deep reader gets that nobody else does]
**What's still opaque:** [what doesn't land even after 20 minutes]
**Return signal:** [would they come back?]

## Transition Hooks
- **10s → 2min:** [present / weak / missing] — [what it is or what's needed]
- **2min → 20min:** [present / weak / missing] — [what it is or what's needed]

## Brittleness Risk
[If the site only works at one speed, which speed? What breaks at the other speeds?]

Direction (if not all passing)

For each BRITTLE or ADEQUATE tier:

What specifically doesn't register at that speed (with evidence)
What would need to change to make it register (specific content or design change)
Which files to modify
What the result should feel like after the change

For missing transition hooks:

What's currently at the transition point
What kind of hook is needed (content, visual, navigational)
Where it should live (which page, which section)

Called By

/full-pass run — parallel arm (independent, no data dependencies)
Standalone — usable anytime