Skill Explosion and Harness Framework Consolidation

Original: Tweet · Kasong (@kasong2048) · March 30, 2026 Engagement: 464 likes · 41 retweets · 68K views · 558 bookmarks Category: Industry Perspective / Practice

The Problem

"Many programmers' Skill counts have gotten completely out of control."

Kasong observed a widespread phenomenon: developers install superpowers today, get talked into compound-engineering tomorrow, and follow the trend with gstack the day after. Skill counts balloon to dozens in no time, and the switching and recall logic between Skills becomes incredibly unstable.

Root Cause

These frameworks are all explorations of Harness Engineering best practices, each with highlights, but with overlapping capabilities. Current Skill management tools take the approach of "hiding complexity" (enable/disable a Skill), rather than "reducing complexity."

Kasong's Solution

Use one framework as the foundation and merge highlights from other frameworks into it, rather than installing them in parallel:

Select base framework (e.g., superpowers)
  → Identify unique highlights from other frameworks (e.g., compound's continuous learning)
    → Use Skills to merge and test other Skills
      → Control the rate of Skill count explosion

Key insight: Use Skills to manage Skills. The merging and testing work itself is also done by the Agent.

Why It Matters

Skill explosion is a systemic problem facing Harness Engineering:

Low discovery cost — Installing a Skill is too easy
High removal cost — You're not sure if removing it will cause problems
Hard interaction testing — The combined behavior of Skill A and Skill B is hard to predict

This is a different manifestation of the same problem described in Zeratul's Law about harness bloat. One occurs at the harness level, the other at the skill level.