lyndonkl/claude/analogy-weight-check

Trust tier 2
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

For every analogy in a substacker draft, verifies it carries mechanical weight — the analogy does real work explaining the mechanism, not merely decorates it. Cross-references analogy-catalog.md for novelty (is this analogy reused from a prior post?) and domain fit (biology > organizational > sports preferred; physics/military disfavored). Use whenever an analogy appears in the draft. Trigger keywords — analogy weight, decorative, mechanical weight, reused analogy, catalog check, metaphor check.

Install

npx vskill@latest install lyndonkl/claude/analogy-weight-check

SKILL.md

Analogy Weight Check

Table of Contents

Related skills: Called by Editor in structural pass (since analogy decisions affect paragraph structure). Reads shared-context/analogy-catalog.md for novelty.

Mechanical-weight test

The test: if you remove the analogy, does the explanation still work?

  • Still works → analogy is decorative. Either cut or extend so it carries weight.
  • Degrades → analogy carries weight. Keep.

A decorative analogy is not a tier-1 issue by itself — writers may legitimately use decorative language occasionally. But:

  • If the draft has >2 decorative analogies, flag as a pattern. Cluster of decorative = slop signal.
  • If the analogy is physics/military AND decorative, flag tier-2 (physics/military are already disfavored directions).

Workflow

For each analogy in the draft:
- [ ] Step 1: Identify the analogy (source + target)
- [ ] Step 2: Mechanical-weight test — simulate removing it; does the explanation degrade?
- [ ] Step 3: Check analogy-catalog.md for reuse
- [ ] Step 4: Check source domain against voice-profile priority (biology > organizational > sports; not physics/military)
- [ ] Step 5: Emit verdict per analogy: carries-weight | decorative | reused-from-catalog | wrong-domain

Worked example

Draft excerpt:

A KV cache is like a library's card catalog — it lets you find things. Under the hood, it stores key and value projections for each past token, so when a new token arrives, attention can index back without recomputing.
Attention is also like a room full of people trying to hear each other.

Analogies:

  1. "KV cache like a library card catalog"
  • Remove test: "A KV cache stores key and value projections for each past token, so when a new token arrives, attention can index back without recomputing." Still works and is arguably clearer.
  • Verdict: decorative. Catalog check: not in analogy-catalog. Domain: neutral (everyday).
  • Suggestion: either cut the simile, OR extend to carry weight ("like a card catalog with a fixed drawer count — evicting old cards to make room for new ones").
  1. "Attention is also like a room full of people trying to hear each other"
  • Remove test: what remains? "Attention is also…" — the sentence collapses; no mechanism.
  • Verdict: decorative AND un-mapped (no component-to-component mapping in the draft).
  • Suggestion: cut entirely or replace with a framing that carries weight.

Overall: 2 decorative analogies in one excerpt. Cluster signal — flag as tier-2, note pattern.

Guardrails

  1. Don't flag an analogy as decorative if the draft immediately extends it (next sentence provides the mapping). Look at the surrounding 2-3 sentences.
  2. An analogy the writer has just generated via Intuition Builder is pre-mapped; treat as carrying weight unless the draft drops the mapping.
  3. Reused analogies from the catalog are not failures; they're a feature of recurring analogies the writer owns (e.g., "from first principles" is a recurring frame). Flag softly.
  4. Physics/military analogies get an extra demerit but are not automatic tier-1 — writer may use them with explicit irony or override.
  5. Suggested rewrites should be ≤2 sentences each. Don't propose a full paragraph.
  6. A decorative analogy in isolation is tier-2. Three+ decorative analogies in a single draft = pattern flag, still tier-2 but surfaced in summary.

Quick reference

  • Per-analogy verdict: carries-weight | decorative | reused-from-catalog | wrong-domain.
  • Cluster rule: >2 decorative = pattern flag.
  • Reads analogy-catalog.md; does not write (that's update-analogy-catalog).

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
155
Repository forks
23
7-day trend score
25

Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 7 Sep 2026, a month ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/lyndonkl/claude/analogy-weight-check/badge)](https://verified-skill.com/skills/lyndonkl/claude/analogy-weight-check)

Something wrong with this skill or its page? Tell us.