Claude Code Plugins

Community-maintained marketplace

Feedback

PARAGON Enforcement System v3.5 - 49 guards for Clean Code, SOLID, configuration centralization, stack compliance, parse-at-boundary, and evidence-based development.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name paragon
description PARAGON Enforcement System v3.5 - 49 guards for Clean Code, SOLID, configuration centralization, stack compliance, parse-at-boundary, and evidence-based development.
allowed-tools Read, Write, Edit, Bash, Grep, Glob
token-budget 500
version 3.5.0
references [object Object], [object Object]

PARAGON Enforcement System v3.5

Protocol for Automated Rules, Analysis, Guards, Observance, and Norms

"The only way to go fast is to go well." — Uncle Bob

Enforcement Layers

Layer Mechanism Trigger
Claude PreToolUse hooks (paragon-guard.ts) Every Write/Edit/Bash
Git pre-commit hooks (git-hooks.nix) Every commit
CI GitHub Actions (paragon-check.yml) Every PR/push

Guard Matrix Summary (49 Guards)

Tier 1: Original Guards (1-14) - BLOCKING

# Guard Blocks
1 Bash Safety rm -rf /, rm -rf ~
2 Conventional Commits Non-conventional messages
3 Forbidden Files package-lock, bun.lock, eslint, jest, .env
4 Forbidden Imports express, prisma, zod/v3, dd-trace
5 Any Type : any, as any, <any>
6 z.infer z.infer<>, z.input<>, z.output<>
7 No-Mock jest.mock, vi.mock
8 TDD Source without test
9 DevOps Files process-compose.yaml, .env
10 DevOps Commands bun run/test/install
13 Assumption Language "should work", "probably"
14 Throw Detector throw for expected errors

Tier 2: Clean Code (15-17)

# Guard Blocks
15 No Comments Unnecessary inline comments
16 Meaningful Names Cryptic abbrevs, Hungarian notation
17 No Commented-Out Code Dead code in comments

Quick Reference (Tiers 3-8)

See references/guards-detail.md for full details on:

  • Tier 3: Extended Clean Code (18-25)
  • Tier 4: Tooling Guards (26-27)
  • Tier 5: Configuration Guards (28-30)
  • Tier 6: Stack Compliance (31)
  • Tier 7: Parse-at-Boundary (32-39)
  • Tier 8: Parse Don't Validate (40-49)

Verification-First Philosophy

BANNED REQUIRED
"should work" "VERIFIED via test: [assertion]"
"this should fix" "UNVERIFIED: requires [test]"
"probably works" "confirmed by running: [cmd]"
"I think" Evidence-based statements

Clean Code Principles Enforced

Principle Guard(s)
Functions should be small 20, 21
Few arguments 18
Meaningful names 16
Don't comment bad code 15, 17
Law of Demeter 19
No null returns 23
Guard clauses over nesting 25

Quick Commands

# Verify PARAGON compliance
just verify-paragon

# Run pre-commit manually
just lint-staged

# Run ast-grep validation
sg scan --rule config/agents/rules/paragon-combined.yaml .

# Skip specific guard (emergency)
touch .paragon-skip-31

Related Skills

Skill Relationship
typescript-patterns Type-first, Result types
effect-ts-patterns Typed errors, Layer DI
tdd-patterns Red-Green-Refactor
hexagonal-architecture No-mock testing
parse-boundary-patterns Guards 32-39 patterns