Claude Code Full Registry Audit

Debate Date: 2026-04-27 | 6-Round Structured Debate

Inventory Summary

Plugins

672

Skills

Agents

Hooks

MCP Servers

227

Commands

~203 REMOVE ~100 MERGE ~350 KEEP ~101 DEPRECATE w/ PATH

Converged Framework: Quality-based curation — populated vs placeholder, production-ready vs dev artifacts, demonstrated utility vs theoretical coverage. All three debaters agreed on: dependency analysis before removal, soft delete over hard delete, phased approach (flag → notify → 90-day grace → remove), and rollback capability within 1 hour.

Round-by-Round Debate Summary

Round 1 — Opening Positions

ADVOCATE

Keep everything. Breadth = coverage, not bloat. Self-regulating ecosystem enables emergence.

SKEPTIC

Remove 60-70%. 51% redundancy is unacceptable. Registry is a dumping ground, not a library.

REALIST

Merge duplicates + deprecate with migration paths. Quality over quantity, but don't destroy user workflows.

Round 2 — Argued (Evidence Exchange)

Key revelations from inventory analysis:

Skeptic found: template-skill (9 placeholders), superpowers-dev (54 artifacts), example-skills (18 entries), test artifacts
Advocate found: Redundancy is 13%, not 51%. Most duplicates are version variants within same plugin.
Realist found: 41 temp_git orphans, 90 placeholder descriptions, 101 user-sourced skills need migration

Round 3 — Defend (Rebuttals)

All three conceded template/placeholder pollution is real. Advocate held firm on domain-specific skills. Skeptic refined from "fewer is better" to "quality determines legitimacy."

Round 4 — Concede (Acknowledgment)

ADVOCATE NOW CONCEDES

"Keep everything" is indefensible. Hollow entries damage registry credibility.

SKEPTIC NOW CONCEDES

Domain-specific skills (Airflow, Shopify, VEX) serve real needs. Version variants serve rollback.

REALIST NOW CONCEDES

Skeptic's pollution identification was accurate. Advocate's specialized coverage defense was correct.

Round 5 — Cross-Examine (Pointed Questions)

Question	Target	Core Challenge
If 9x frontend-design copies are identical, which one do you keep?	Skeptic	Quality standards collapse without selection criteria
How distinguish template from early-stage skill?	Advocate	Any quantitative threshold is a proxy for judgment
What is actual maintenance cost (in hours)?	Both	Neither side has real numbers; pure estimation
What is rollback plan if removal breaks something?	Both	No reversibility mechanism proposed

Round 6 — Synthesis (Final Recommendations)

Operational Requirements (All Three Agree):
1. Dependency graph analysis before any removal
2. Soft delete over hard delete — archive, don't delete
3. Phased approach: flag → notify authors → 90-day grace → remove unresponsive
4. Rollback capability within 1 hour
5. Schema validation at submission to prevent hollow entries

Plugins (22 total)

KEEP

compound-engineering

cache (v2.68.1)

AI-powered development tools for code review, research, design, and workflow automation

MERGE

autoresearch

cache + marketplace

Autonomous improvement engine — exists in both locations, consolidate versions

MERGE

understand-anything

cache + marketplace

AI-powered codebase understanding — duplicate entries across sources

REMOVE

invalid-command-path

marketplace

Empty description — broken plugin entry

KEEP

claude-mem

cache (v10.4.1)

Persistent memory system for Claude Code

KEEP

claudeclaw

cache (v1.0.0)

Cron-like daemon for scheduled prompts

KEEP

ui-ux-pro-max

cache (v2.0.1)

UI/UX design intelligence — 50 styles, 21 palettes, 50 font pairings

Skills (672 total — Key Verdicts)

REMOVE

template-skill

9 duplicates

"Replace with description" — placeholder scaffolding, never meant for production

REMOVE

temp_git* orphans

41 entries

Orphaned git clone artifacts — should have been garbage collected

REMOVE

example-skills/*

18 entries

Developer scaffolding left in production

REMOVE

superpowers-dev/*

54 entries

Development subdirectory artifacts never cleaned on plugin update

MERGE

frontend-design

11 duplicates → 1

Identical entries across plugin builds — consolidate to latest

MERGE

skill-creator

11 duplicates → 1

Same skill across multiple plugin versions

MERGE

algorithmic-art

9 duplicates → 1

Canvas design variants — example-skills artifacts

MERGE

docx / pdf / pptx / xlsx

9x each → 1 each

Brand-style skill family — identical copies

KEEP

airflow family

16+ specialized skills

Airflow, airflow-plugins, deploying-airflow, debugging-dags, authoring-dags, etc.

KEEP

shopify-* family

20+ specialized skills

Admin API, Storefront GraphQL, Polaris UI, Liquid, Hydrogen, Customer accounts

KEEP

vex-robotics-cpp-expert

specialized domain

VEX robotics C++ expertise for competition programming

KEEP

tax-expert

specialized domain

Tax calculation and SEC filing logic

KEEP

cloudflare:* skills

cloudflare plugin

Workers, Pages, R2, D1, Durable Objects, Vectorize

KEEP

pm-* family

pm-execution, pm-toolkit, pm-product-strategy, etc.

Product management comprehensive toolkit

DEPRECATE w/ PATH

user-sourced skills

101 entries

Workflow dependencies exist — 90-day grace period for author response

Agents (56 total)

KEEP

compound-engineering:* agents

compound-engineering plugin

Code review, feature-dev, pr-review-toolkit, plugin-dev, understand-anything

REMOVE

temp_git* agents

test artifacts

Orphaned git clone artifacts — garbage collection needed

Hooks (29 total)

MERGE

session-start hooks

4 duplicates

Same trigger across plugin versions — consolidate to latest

MERGE

pretooluse hooks

3 duplicates

Same trigger across versions

REMOVE

test fixture hooks

echo hooks

echo before, echo custom, echo default — test fixtures posing as hooks

MCP Servers (8 total)

MERGE

chrome-devtools-mcp

3 duplicates

Same server across versions (0.22.0, latest, temp_git) — consolidate to latest

MERGE

episodic-memory

2 duplicates

Same memory server across locations

REMOVE

example-server

test fixture

Dev artifact — clearly not for production

KEEP

aws-serverless

production

AWS Lambda, API Gateway, SAM deployment

KEEP

cloudflare

production

Workers, Pages, D1, Durable Objects

Commands (227 total)

MERGE

hookify:* commands

5x each across plugin versions

Same commands triplicated — consolidate

MERGE

claudeclaw:* commands

3x each

Same commands across plugin versions

REMOVE

help command duplicates

8 instances

Same help command across versions

KEEP

unique commands

~60 unique functions

Deduplicated command set

Final Actionable Recommendations

Phase	Action	Estimated Scope
Immediate	Remove template placeholders, temp_git orphans, example-skills, superpowers-dev, placeholder descriptions	~203 entries
Phase 1	Merge frontend-design (11→1), algorithmic-art (9→1), skill-creator (11→1), brand-style skills	~100 entries → ~20
Phase 2	Flag user-sourced skills, notify authors, 90-day grace period	~101 entries
Infrastructure	Build dependency graph, soft delete system, rollback capability, schema validation	Engineering investment

Debate Participants

Advocate: Argued for breadth and optionality — evolved to support quality-based curation

Skeptic: Argued for aggressive removal — evolved to focus on execution quality, not quantity

Realist: Argued for pragmatic middle ground — provided the converged framework