Two files, one repository

carrot-foundation/middle-earth ships 3 formats across 13 indexed files. The question worth asking is whether the second one says anything the first does not.

CompareAGENTS.md ↔ CLAUDE.md AGENTS.md ↔ Cursor rules CLAUDE.md ↔ Cursor rules

A · AGENTS.md · 1105 words B · CLAUDE.md · 780 words

What each file covers, counted
Dimension	Shared	Only in A	Only in B	Overlap
Sections	35	8	4	74%
Commands	16	0	0	100%
Section tags	10	1	0	91%

What each file covers

Sections

35 shared · 8 only in A · 4 only in B

− AGENTS.md
− Canonical workflow
− Current capability counts
− Available skills
− Rule mappings
− Agent roles
− Runtime adapter paths
− Setup commands (workspace)
+ CLAUDE.md
+ Claude runtime
+ Required workflow
+ Capability counts
Equality rule
Canonical references
Middle Earth Project Context
Project Overview
Technology Stack
Architecture
Directory Structure
Path Aliases
UI Library (web-ui)
Key Constraints
Common Commands
Development
Build a library
Run tests
Run linting
Affected Commands
Lint affected projects
Test affected projects (not available — use nx directly)
Build affected projects (not available — use nx directly)
Utilities
Commitizen (guided commit message)
Validate package.json files
Spell check
Code Standards
TypeScript
Testing
Commit Message Guidelines
Quick Reference
Middle Earth Scopes
Examples
Nx Configuration
Important Notes
Troubleshooting
Clear Caches
Dependency Issues

Commands

16 shared · 0 only in A · 0 only in B

eslint-config-carrot/
pnpm nx build <project-name>
pnpm nx test <project-name>
pnpm nx lint <project-name>
pnpm lint:affected
pnpm nx affected --target test
pnpm nx affected --target build
pnpm commit
pnpm pkgJsonLint
pnpm spellcheck
pnpm nx reset
pnpm install
pnpm ai:sync
pnpm ai:check
jest.mock
jest.spyOn

Section tags

10 shared · 1 only in A · 0 only in B

− monorepo
setup
build
test
lint-format
architecture
types
testing-strategy
git-pr
ui
agent-behaviour

Line diff

+17 added−55 removed201 unchanged78.5% identical

carrot-foundation/middle-earth · AGENTS.md

@@ −1 @@

1# AGENTS.md

3Middle Earth AI instructions for Codex, Claude, and Cursor with equal capability parity.

5## Equality rule

7- Cursor, Claude, and Codex are treated as equals.

8- No platform is primary for instruction definition.

9- Canonical source: `.ai/`.

11## Canonical workflow

131. Edit canonical files in `.ai/`.

142. Run `pnpm ai:sync` to regenerate platform adapters.

153. Run `pnpm ai:check` to validate parity and links.

17## Current capability counts

19- Rules: 11

20- Skills: 8

21- Agents/Roles: 2

23## Available skills

25- `check` - Run Middle Earth quality gates (lint, test, build) without committing. Runs AI instruction checks when .ai/ files are affected.

26- `commit` - Commit Changes

27- `create-branch` - Create Branch

28- `create-pr` - Create Pull Request

29- `debug` - Structured debugging workflow for errors, test failures, build issues, and unexpected behavior in the Middle Earth monorepo.

30- `finish-work` - End-to-end workflow to ship code changes as a PR. Creates branch, runs quality gates, commits changes, and creates PR.

31- `review-pr` - Review pull requests for quality, correctness, and adherence to Middle Earth project standards.

32- `unit-test` - Run and write Jest unit tests following Middle Earth patterns.

34## Rule mappings

36- `rule-bmad` - BMAD tool/output split and per-app artifact convention for Middle Earth

37- `rule-branch-naming` - Git branch naming conventions aligned with Conventional Commits

38- `rule-code-comments` - Code comment guidelines for all languages

39- `rule-code-preservation` - Code preservation rules - never delete or revert code without explicit approval

40- `rule-code-style` - Middle Earth code style conventions (TypeScript/JavaScript)

41- `rule-commit` - Conventional Commit message standards with Middle Earth scopes

42- `rule-documentation` - Markdown documentation standards

43- `rule-pull-request` - Pull request workflow and quality standards

44- `rule-pull-request-description` - Pull request description format and content standards

45- `rule-testing` - Middle Earth testing baseline guidelines (unit tests with Jest)

46- `rule-typescript` - Middle Earth TypeScript baseline guidelines (compiler, modules, best practices)

48## Agent roles

50- `debugger` - Debugging specialist for errors, test failures, and unexpected behavior. Use proactively when you hit failures that need root-cause analysis.

51- `verifier` - Validates completed work. Use after tasks are marked done to confirm implementations are functional and meet project standards.

53## Canonical references

55- `.ai/README.md`

@@ −58 @@

58- `.ai/PARITY_MATRIX.md`

59- `.ai/PROJECT_CONTEXT.md`

61## Runtime adapter paths

63- Cursor: `.cursor/rules/`, `.cursor/skills/`, `.cursor/agents/`

64- Claude: `.claude/settings.json`, `.claude/skills/`, `.claude/agents/`

65- Codex: `.agents/skills/`, `AGENTS.md`

67## Setup commands (workspace)

69- Install deps: `pnpm install`

70- Run lint (affected): `pnpm lint:affected`

71- Run tests: `pnpm nx affected --target test`

72- Build: `pnpm nx affected --target build`

73- Validate AI instructions: `pnpm ai:check`

75# Middle Earth Project Context

carrot-foundation/middle-earth · CLAUDE.md

@@ +1 @@

1# CLAUDE.md

3Claude adapter for Middle Earth AI instructions. This file is generated from canonical `.ai/`.

5## Equality rule

7- Cursor, Claude, and Codex are configured as equals.

8- Capability parity is mandatory across all three.

9- Canonical source remains tool-agnostic in `.ai/`.

11## Claude runtime

13- Baseline settings: `.claude/settings.json`

14- Skills: `.claude/skills/*/SKILL.md`

15- Agents: `.claude/agents/*.md`

17## Required workflow

191. Update canonical docs under `.ai/`.

202. Run `pnpm ai:sync`.

213. Run `pnpm ai:check`.

23## Canonical references

25- `.ai/README.md`

@@ +28 @@

28- `.ai/PARITY_MATRIX.md`

29- `.ai/PROJECT_CONTEXT.md`

31## Capability counts

33- Rules: 11

34- Skills: 8

35- Agents/Roles: 2

37# Middle Earth Project Context

@@ −1 +1 @@

1−# AGENTS.md

1+# CLAUDE.md

3−Middle Earth AI instructions for Codex, Claude, and Cursor with equal capability parity.

3+Claude adapter for Middle Earth AI instructions. This file is generated from canonical `.ai/`.

55 ## Equality rule

7−- Cursor, Claude, and Codex are treated as equals.

8−- No platform is primary for instruction definition.

9−- Canonical source: `.ai/`.

7+- Cursor, Claude, and Codex are configured as equals.

8+- Capability parity is mandatory across all three.

9+- Canonical source remains tool-agnostic in `.ai/`.

1010

11−## Canonical workflow

11+## Claude runtime

1212

13−1. Edit canonical files in `.ai/`.

14−2. Run `pnpm ai:sync` to regenerate platform adapters.

15−3. Run `pnpm ai:check` to validate parity and links.

13+- Baseline settings: `.claude/settings.json`

14+- Skills: `.claude/skills/*/SKILL.md`

15+- Agents: `.claude/agents/*.md`

1616

17−## Current capability counts

17+## Required workflow

1818

19−- Rules: 11

20−- Skills: 8

21−- Agents/Roles: 2

19+1. Update canonical docs under `.ai/`.

20+2. Run `pnpm ai:sync`.

21+3. Run `pnpm ai:check`.

2222

23−## Available skills

24−

25−- `check` - Run Middle Earth quality gates (lint, test, build) without committing. Runs AI instruction checks when .ai/ files are affected.

26−- `commit` - Commit Changes

27−- `create-branch` - Create Branch

28−- `create-pr` - Create Pull Request

29−- `debug` - Structured debugging workflow for errors, test failures, build issues, and unexpected behavior in the Middle Earth monorepo.

30−- `finish-work` - End-to-end workflow to ship code changes as a PR. Creates branch, runs quality gates, commits changes, and creates PR.

31−- `review-pr` - Review pull requests for quality, correctness, and adherence to Middle Earth project standards.

32−- `unit-test` - Run and write Jest unit tests following Middle Earth patterns.

33−

34−## Rule mappings

35−

36−- `rule-bmad` - BMAD tool/output split and per-app artifact convention for Middle Earth

37−- `rule-branch-naming` - Git branch naming conventions aligned with Conventional Commits

38−- `rule-code-comments` - Code comment guidelines for all languages

39−- `rule-code-preservation` - Code preservation rules - never delete or revert code without explicit approval

40−- `rule-code-style` - Middle Earth code style conventions (TypeScript/JavaScript)

41−- `rule-commit` - Conventional Commit message standards with Middle Earth scopes

42−- `rule-documentation` - Markdown documentation standards

43−- `rule-pull-request` - Pull request workflow and quality standards

44−- `rule-pull-request-description` - Pull request description format and content standards

45−- `rule-testing` - Middle Earth testing baseline guidelines (unit tests with Jest)

46−- `rule-typescript` - Middle Earth TypeScript baseline guidelines (compiler, modules, best practices)

47−

48−## Agent roles

49−

50−- `debugger` - Debugging specialist for errors, test failures, and unexpected behavior. Use proactively when you hit failures that need root-cause analysis.

51−- `verifier` - Validates completed work. Use after tasks are marked done to confirm implementations are functional and meet project standards.

52−

5323 ## Canonical references

5424

5525 - `.ai/README.md`

@@ −58 +28 @@

5828 - `.ai/PARITY_MATRIX.md`

5929 - `.ai/PROJECT_CONTEXT.md`

6030

61−## Runtime adapter paths

31+## Capability counts

6232

63−- Cursor: `.cursor/rules/`, `.cursor/skills/`, `.cursor/agents/`

64−- Claude: `.claude/settings.json`, `.claude/skills/`, `.claude/agents/`

65−- Codex: `.agents/skills/`, `AGENTS.md`

66−

67−## Setup commands (workspace)

68−

69−- Install deps: `pnpm install`

70−- Run lint (affected): `pnpm lint:affected`

71−- Run tests: `pnpm nx affected --target test`

72−- Build: `pnpm nx affected --target build`

73−- Validate AI instructions: `pnpm ai:check`

33+- Rules: 11

34+- Skills: 8

35+- Agents/Roles: 2

7436

7537 # Middle Earth Project Context

7638

Two files, one repository

Built for the same person as RuleStack