← Back to Developer Overview

Helpfulness Scoring

Autonomous Depth Decisions: How the skill calculates which topics deserve their own page

01

What is the Helpfulness Score?

The Helpfulness Score (HS) is a value from 1 to 10 that the skill calculates for every potential sub-topic. It answers one central question: Does this topic deserve its own page, or should it remain a paragraph on the parent page?

The score is calculated per topic and per audience. The same topic might get a dedicated detail page for developers while remaining just a brief paragraph for end users.

HS = C + R + L + I
0–3
Complexity (C)
0–3
Relevance (R)
0–2
Learning Value (L)
0–2
Independence (I)

Maximum score: 3 + 3 + 2 + 2 = 10

Why not give every topic its own page? Because information overload destroys learning effectiveness. A topic with HS 2 as a standalone page would contain barely a paragraph and slow the reader down with unnecessary navigation. The HS ensures depth is created only where it delivers real value.

02

The Four Dimensions

Each dimension is evaluated independently. Their sum yields the HS. Here are the exact scoring criteria:

Dimension 0 1 2 3
Complexity
How much explanation does the topic need?
Trivial, self-explanatory Simple, one paragraph suffices Medium, needs examples High, needs own page with visuals
Relevance
How important is it for the audience?
Niche topic Nice to know Important for understanding Core concept, essential
Learning Value
Does it change understanding?
Plain facts, look-up-able Useful knowledge Fundamentally changes understanding
Independence
Does it stand on its own?
Repeats parent page Supplements parent page Completely standalone topic
🔍
Example: OAuth Flow
Complexity 3 (multiple steps, tokens, redirects), Relevance 3 (core feature), Learning Value 2 (changes security understanding), Independence 2 (standalone system)
HS = 10
📄
Example: License Info
Complexity 0 (trivial), Relevance 1 (nice to know), Learning Value 0 (plain facts), Independence 0 (repeats README)
HS = 1
03

Audience-Specific Thresholds

The same score has different consequences for different audiences. Developers get their own pages faster because technical depth is expected. Executives rarely get detail pages because they seek overview, not depth.

Thresholds by Audience

Audience Own Page at Deeper Level at Example: “OAuth Implementation”
🔧 Developers ≥ 6 ≥ 8 HS 9 → Own page + deeper level
👤 Users ≥ 7 ≥ 9 HS 3 → Paragraph on parent page only
📊 Executives ≥ 8 ≥ 10 HS 2 → Not even mentioned

Why are the thresholds different? Developers expect technical depth — they actively seek details. Users want to accomplish tasks — too many pages create confusion. Executives need overview — every detail page is potentially one too many. The threshold of ≥ 10 for executives at the deeper level means: practically never.

✍️ Quick Check

The topic “Token Refresh” has: Complexity=2, Relevance=3, LearningValue=2, Independence=2. What is the HS? Does it get its own page for 👤 Users?

HS = 7 — No, below the threshold for users
HS = 9 — No, users don't need technical details
HS = 9 — Yes, because ≥ 7 is enough for a user page
HS = 10 — Yes, maximum score
04

The Decision Tree

For each topic T, audience A, and level L, the skill walks through the following decision tree. The result is always one of four actions.

Evaluate Topic T for Audience A at Level L
Step 1
L ≥ max_level?
Yes
STOP — No deeper level possible
No → Step 2
HS < threshold(A)?
Yes
Paragraph on parent page — No own page
No → Step 3
HS ≥ threshold(A) AND HS < deeper_threshold(A)?
Yes
Create own page — No deeper level
No → Step 4
HS ≥ deeper_threshold(A) AND L+1 < max_level?
Yes
Create own page + Plan next level
function decide(T, A, L):
    // Step 1: Check level boundary
    if L >= max_level(A):
        return STOP

    hs = calculate_hs(T, A)

    // Step 2: Below threshold?
    if hs < threshold(A):
        return PARAGRAPH

    // Step 3: Own page, but no deeper
    if hs < deeper_threshold(A):
        create_page(T, A, L)
        return PAGE_NO_DEPTH

    // Step 4: Own page + next level
    if L + 1 < max_level(A):
        create_page(T, A, L)
        plan_next_level(T, A, L + 1)
        return PAGE_WITH_DEPTH
    else:
        create_page(T, A, L)
        return PAGE_NO_DEPTH
For each topic T, audience A, level L:

1. Is the current level already the maximum for this audience?
   → Yes: Stop. No deeper level possible.

2. Calculate the Helpfulness Score for T relative to A.
   Is the HS below the threshold for A?
   → Yes: Topic is treated as a paragraph on the parent page.

3. Is the HS high enough for its own page,
   but below the "deeper" threshold?
   → Yes: Create own page, but no further depth.

4. Is the HS above the "deeper" threshold
   AND is the next level still below the maximum?
   → Yes: Create own page AND plan the next level.
   → Otherwise: Own page without further depth.
05

Typical HS Patterns

Certain topic types reliably land in specific score ranges. These patterns show how the HS shapes content structure for each audience.

HS Heatmap by Topic Type and Audience

Topic Type 🔧 Developers 👤 Users 📊 Executives
Implementation Details
Code architecture, API internals, algorithms
7–10 1–3 1–3
UI Workflows
User interface, step-by-step guides
3–5 7–10 3–5
Cost / ROI
Budget, economics, business case
1–3 2–4 8–10
Architecture Decisions
Design patterns, trade-offs, tech stack
7–9 1–3 6–8

Spot the pattern: Implementation details and architecture decisions are almost always relevant for developers (HS 7+), but rarely for users (HS 1–3). Conversely, UI workflows dominate for users. Cost/ROI is the executives' domain. Architecture decisions are the only category that generates deeper pages for both developers and executives — because trade-offs concern both groups.

Deep Dive Score Calculation and Weighting → How the score is internally computed, which weights apply, and how edge cases are handled.
More developer deep-dive pages