Two files, one repository

caliber-ai-org/ai-setup ships 3 formats across 6 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 · 442 words B · CLAUDE.md · 783 words

What each file covers, counted
Dimension	Shared	Only in B	Overlap
Sections	6	3	67%
Commands	8	2	80%
Section tags	5	0	100%

What each file covers

Sections

6 shared · 0 only in A · 3 only in B

+ What Is This
+ Key Patterns
+ Context Sync
Caliber
Commands
Architecture
Conventions
Before Committing
Session Learnings

Commands

8 shared · 0 only in A · 2 only in B

+ npx vitest run src/scoring/__tests__/accuracy.test.ts
+ git.ts
npm run build
npm run dev
npm run test
npm run lint
npx tsc --noEmit
vitest.config.ts
eslint.config.js
git-diff.ts

Section tags

5 shared · 0 only in A · 0 only in B

build
test
lint-format
code-style
git-pr

Line diff

+53 added−19 removed52 unchanged49.5% identical

caliber-ai-org/ai-setup · AGENTS.md

@@ −1 @@

1# Caliber

3`@rely-ai/caliber` — AI context infrastructure CLI. Scores, generates, and syncs agent configs.

5## Commands

7```bash

@@ −10 @@

10npm run test # vitest run

11npm run lint # eslint src/

12npx tsc --noEmit # type check

13```

15## Architecture

17**Entry**: `src/bin.ts` → `src/cli.ts` · **Build**: `tsup.config.ts` · **Test**: `vitest.config.ts` · **Lint**: `eslint.config.js`

19**Commands** (`src/commands/`): `init.ts` · `score.ts` · `refresh.ts` · `regenerate.ts` · `config.ts` · `hooks.ts` · `learn.ts` · `recommend.ts` · `sources.ts` · `undo.ts` · `status.ts` · `setup-files.ts`

21**LLM** (`src/llm/`): `anthropic.ts` · `vertex.ts` · `openai-compat.ts` · `cursor-acp.ts` · `claude-cli.ts` · `types.ts` · `config.ts` · `model-recovery.ts`

23**AI** (`src/ai/`): `generate.ts` · `refine.ts` · `refresh.ts` · `detect.ts` · `learn.ts` · `score-refine.ts` · `prompts.ts`

25**Fingerprint** (`src/fingerprint/`): `index.ts` · `file-tree.ts` · `code-analysis.ts` · `sources.ts` · `cache.ts`

27**Scoring** (`src/scoring/`): `index.ts` · `constants.ts` · `display.ts` · Checks: `checks/existence.ts` · `checks/quality.ts` · `checks/grounding.ts` · `checks/accuracy.ts` · `checks/freshness.ts` · `checks/bonus.ts`

29**Writers** (`src/writers/`): `index.ts` · `claude/index.ts` · `cursor/index.ts` · `codex/index.ts` · `github-copilot/index.ts` · `staging.ts` · `manifest.ts` · `backup.ts`

31**Lib** (`src/lib/`): `hooks.ts` · `state.ts` · `resolve-caliber.ts` · `builtin-skills.ts` · `sanitize.ts` · `git-diff.ts`

33**Other**: `src/constants.ts` · `src/test/setup.ts` · `github-action/` · `packages/mcp-server/` · `packages/shared/` · `apps/`

35## Conventions

37- ESM with `.js` extensions · Tests in `__tests__/` dirs · `unknown` over `any`

38- Conventional commits (`feat:`, `fix:`, `refactor:`) · Dev: `next`, Stable: `master`

39- Providers implement `LLMProvider` from `src/llm/types.ts`

40- Writers in `src/writers/claude/index.ts` · `src/writers/codex/index.ts` return `string[]`

41- Scoring checks in `src/scoring/checks/` return `Check[]`

42- No hardcoded mappings — LLM-driven via `src/ai/detect.ts`

43- `collectFingerprint()` async with LLM enrichment (`src/fingerprint/index.ts`)

44- MCP server workspace: `packages/mcp-server/`

46

47## Before Committing

@@ −56 @@

56- If **hook-active**: the hook handles sync automatically — just commit normally. Tell the user: "Caliber will sync your agent configs automatically via the pre-commit hook."

57- If **no-hook**: run Caliber manually before committing:

58 1. Tell the user: "Caliber: Syncing agent configs with your latest changes..."

59 2. Run: `caliber refresh && git add CLAUDE.md .claude/ .cursor/ .cursorrules .github/copilot-instructions.md .github/instructions/ AGENTS.md CALIBER_LEARNINGS.md 2>/dev/null`

60 3. After it completes, briefly tell the user what Caliber updated. Then proceed with the commit.

62If `caliber` is not found, tell the user: "This project uses Caliber for agent config sync. Run /setup-caliber to get set up."

@@ −68 @@

68Read `CALIBER_LEARNINGS.md` for patterns and anti-patterns learned from previous sessions.

69These are auto-extracted from real tool usage — treat them as project-specific rules.

70

caliber-ai-org/ai-setup · CLAUDE.md

@@ +1 @@

1# Caliber

3## What Is This

5`@rely-ai/caliber` — CLI that keeps AI agent configs in sync with your codebase. Generates and refreshes `CLAUDE.md`, `.cursor/rules/`, `AGENTS.md`, `.github/copilot-instructions.md`, and skills across Claude Code, Cursor, Codex, OpenCode, and GitHub Copilot. Supports Anthropic, OpenAI, Google Vertex AI, OpenAI-compatible endpoints, Claude Code CLI, and Cursor ACP.

7## Commands

9```bash

@@ +12 @@

12npm run test # vitest run

13npm run lint # eslint src/

14npx tsc --noEmit # type check

15npx vitest run src/scoring/__tests__/accuracy.test.ts # single test

16```

18## Architecture

20**Entry**: `src/bin.ts` → `src/cli.ts` (Commander.js) · **Config**: `tsconfig.json` · `tsup.config.ts` · `vitest.config.ts` · `eslint.config.js` · `.prettierrc`

22**Commands** (`src/commands/`): `init.ts` · `score.ts` · `refresh.ts` · `regenerate.ts` · `config.ts` · `hooks.ts` · `insights.ts` · `learn.ts` · `recommend.ts` · `sources.ts` · `publish.ts` · `undo.ts` · `status.ts` · `bootstrap.ts` · `uninstall.ts` · Helpers: `init-helpers.ts` · `init-prompts.ts` · `init-display.ts` · `setup-files.ts` · `interactive-provider-setup.ts`

24**LLM** (`src/llm/`): `anthropic.ts` · `vertex.ts` · `openai-compat.ts` · `cursor-acp.ts` · `claude-cli.ts` · `types.ts` · `config.ts` · `utils.ts` · `usage.ts` · `model-recovery.ts` · `seat-based-errors.ts` · `index.ts`

26**AI** (`src/ai/`): `generate.ts` · `refine.ts` · `refresh.ts` · `detect.ts` · `learn.ts` · `score-refine.ts` · `prompts.ts` · `stream-parser.ts` · `index.ts`

28**Fingerprint** (`src/fingerprint/`): `index.ts` · `file-tree.ts` · `code-analysis.ts` · `existing-config.ts` · `sources.ts` · `git.ts` · `cache.ts`

30**Scoring** (`src/scoring/`): `index.ts` · `display.ts` · `constants.ts` · `utils.ts` · `history.ts` · `dismissed.ts` · Checks (`src/scoring/checks/`): `existence.ts` · `quality.ts` · `grounding.ts` · `accuracy.ts` · `freshness.ts` · `bonus.ts` · `sources.ts`

32**Writers** (`src/writers/`): `index.ts` · `claude/index.ts` · `cursor/index.ts` · `codex/index.ts` · `opencode/index.ts` · `github-copilot/index.ts` · `refresh.ts` · `staging.ts` · `backup.ts` · `manifest.ts` · `pre-commit-block.ts`

34**Scanner** (`src/scanner/`): `index.ts` — detects local `.mcp.json`, `.cursor/mcp.json` MCP servers, rules, and skills across platforms

36**Lib** (`src/lib/`): `hooks.ts` · `learning-hooks.ts` · `state.ts` · `resolve-caliber.ts` · `builtin-skills.ts` · `sanitize.ts` · `notifications.ts` · `git-diff.ts` · `lock.ts` · `debug-report.ts` · `config-discovery.ts` · `terminal.ts`

38**Utils** (`src/utils/`): `parallel-tasks.ts` · `spinner-messages.ts` · `editor.ts` · `review.ts` · `prompt.ts` · `version-check.ts` · `dependencies.ts` · `waiting-content.ts` · `waiting-cards.json`

40**Telemetry** (`src/telemetry/`): `index.ts` · `config.ts` · `events.ts` · **Learner** (`src/learner/`): `writer.ts` · `storage.ts` · `attribution.ts` · `roi.ts` · `utils.ts` · `stdin.ts`

42**Other**: `github-action/action.yml` · `github-action/index.js` · `assets/video/` (Remotion) · `scripts/` · `docs/FLOW.md` · `src/constants.ts` · `src/test/setup.ts` · `CHANGELOG.md` · `TODOS.md`

44**Workspaces**: `packages/shared/` · `packages/mcp-server/` (MCP server) · `apps/web/` · `apps/api/`

46@./CONTRIBUTING.md

48## Conventions

50- ESM with `.js` import extensions · Tests in `__tests__/` dirs · Setup: `src/test/setup.ts`

51- `unknown` over `any` · Conventional commits (`feat:`, `fix:`, `refactor:`)

52- Dev: `next` branch · Stable: `master` · Node >= 20

53- Config: `~/.caliber/config.json` (mode `0600`) · Constants: `src/constants.ts`

54- `package.json` scripts: `build`, `dev`, `test`, `lint`, `lint:fix`, `format`, `format:check`

56## Key Patterns

58- **Providers**: implement `LLMProvider` from `src/llm/types.ts` (`call()`, `stream()`)

59- **Writers**: `src/writers/claude/index.ts` · `src/writers/cursor/index.ts` · `src/writers/codex/index.ts` · `src/writers/opencode/index.ts` each return `string[]`

60- **Scoring**: checks in `src/scoring/checks/` return `Check[]`, constants in `src/scoring/constants.ts`

61- **No hardcoded mappings**: detection is LLM-driven via `src/ai/detect.ts`

62- `collectFingerprint()` in `src/fingerprint/index.ts` — async with LLM enrichment

63- `resolveCaliber()` from `src/lib/resolve-caliber.ts` for CLI binary resolution

64- Cursor provider: `agent --print --trust --workspace /tmp` in `src/llm/cursor-acp.ts`

65- `getFastModel()` in `src/llm/config.ts`: `ANTHROPIC_SMALL_FAST_MODEL` scoped to anthropic/vertex

66- `validateModel()` skips seat-based providers (`isSeatBased()` in `src/llm/types.ts`)

67- Scoring: deterministic, no LLM · Backups via `src/writers/backup.ts`

68- Scanner: `src/scanner/index.ts` detects local MCP servers, rules, skills for state comparison

69- MCP server workspace: `packages/mcp-server/` · Config discovery: `src/lib/config-discovery.ts`

70- Builtin skills (`src/lib/builtin-skills.ts`): `find-skills`, `save-learning`, `setup-caliber`

72

73## Before Committing

@@ +82 @@

82- If **hook-active**: the hook handles sync automatically — just commit normally. Tell the user: "Caliber will sync your agent configs automatically via the pre-commit hook."

83- If **no-hook**: run Caliber manually before committing:

84 1. Tell the user: "Caliber: Syncing agent configs with your latest changes..."

85 2. Run: `caliber refresh && git add CLAUDE.md .claude/ .cursor/ .cursorrules .github/copilot-instructions.md .github/instructions/ AGENTS.md CALIBER_LEARNINGS.md .agents/ .opencode/ 2>/dev/null`

86 3. After it completes, briefly tell the user what Caliber updated. Then proceed with the commit.

88If `caliber` is not found, tell the user: "This project uses Caliber for agent config sync. Run /setup-caliber to get set up."

@@ +94 @@

94Read `CALIBER_LEARNINGS.md` for patterns and anti-patterns learned from previous sessions.

95These are auto-extracted from real tool usage — treat them as project-specific rules.

96

98

99## Context Sync

100

101This project uses [Caliber](https://github.com/caliber-ai-org/ai-setup) to keep AI agent configs in sync across Claude Code, Cursor, Copilot, and Codex.

102Configs update automatically before each commit via `caliber refresh`.

103If the pre-commit hook is not set up, run `/setup-caliber` to configure everything automatically.

104

105

@@ −1 +1 @@

11 # Caliber

3−`@rely-ai/caliber` — AI context infrastructure CLI. Scores, generates, and syncs agent configs.

3+## What Is This

5+`@rely-ai/caliber` — CLI that keeps AI agent configs in sync with your codebase. Generates and refreshes `CLAUDE.md`, `.cursor/rules/`, `AGENTS.md`, `.github/copilot-instructions.md`, and skills across Claude Code, Cursor, Codex, OpenCode, and GitHub Copilot. Supports Anthropic, OpenAI, Google Vertex AI, OpenAI-compatible endpoints, Claude Code CLI, and Cursor ACP.

57 ## Commands

79 ```bash

@@ −10 +12 @@

1012 npm run test # vitest run

1113 npm run lint # eslint src/

1214 npx tsc --noEmit # type check

15+npx vitest run src/scoring/__tests__/accuracy.test.ts # single test

1316 ```

1417

1518 ## Architecture

1619

17−**Entry**: `src/bin.ts` → `src/cli.ts` · **Build**: `tsup.config.ts` · **Test**: `vitest.config.ts` · **Lint**: `eslint.config.js`

20+**Entry**: `src/bin.ts` → `src/cli.ts` (Commander.js) · **Config**: `tsconfig.json` · `tsup.config.ts` · `vitest.config.ts` · `eslint.config.js` · `.prettierrc`

1821

19−**Commands** (`src/commands/`): `init.ts` · `score.ts` · `refresh.ts` · `regenerate.ts` · `config.ts` · `hooks.ts` · `learn.ts` · `recommend.ts` · `sources.ts` · `undo.ts` · `status.ts` · `setup-files.ts`

22+**Commands** (`src/commands/`): `init.ts` · `score.ts` · `refresh.ts` · `regenerate.ts` · `config.ts` · `hooks.ts` · `insights.ts` · `learn.ts` · `recommend.ts` · `sources.ts` · `publish.ts` · `undo.ts` · `status.ts` · `bootstrap.ts` · `uninstall.ts` · Helpers: `init-helpers.ts` · `init-prompts.ts` · `init-display.ts` · `setup-files.ts` · `interactive-provider-setup.ts`

2023

21−**LLM** (`src/llm/`): `anthropic.ts` · `vertex.ts` · `openai-compat.ts` · `cursor-acp.ts` · `claude-cli.ts` · `types.ts` · `config.ts` · `model-recovery.ts`

24+**LLM** (`src/llm/`): `anthropic.ts` · `vertex.ts` · `openai-compat.ts` · `cursor-acp.ts` · `claude-cli.ts` · `types.ts` · `config.ts` · `utils.ts` · `usage.ts` · `model-recovery.ts` · `seat-based-errors.ts` · `index.ts`

2225

23−**AI** (`src/ai/`): `generate.ts` · `refine.ts` · `refresh.ts` · `detect.ts` · `learn.ts` · `score-refine.ts` · `prompts.ts`

26+**AI** (`src/ai/`): `generate.ts` · `refine.ts` · `refresh.ts` · `detect.ts` · `learn.ts` · `score-refine.ts` · `prompts.ts` · `stream-parser.ts` · `index.ts`

2427

25−**Fingerprint** (`src/fingerprint/`): `index.ts` · `file-tree.ts` · `code-analysis.ts` · `sources.ts` · `cache.ts`

28+**Fingerprint** (`src/fingerprint/`): `index.ts` · `file-tree.ts` · `code-analysis.ts` · `existing-config.ts` · `sources.ts` · `git.ts` · `cache.ts`

2629

27−**Scoring** (`src/scoring/`): `index.ts` · `constants.ts` · `display.ts` · Checks: `checks/existence.ts` · `checks/quality.ts` · `checks/grounding.ts` · `checks/accuracy.ts` · `checks/freshness.ts` · `checks/bonus.ts`

30+**Scoring** (`src/scoring/`): `index.ts` · `display.ts` · `constants.ts` · `utils.ts` · `history.ts` · `dismissed.ts` · Checks (`src/scoring/checks/`): `existence.ts` · `quality.ts` · `grounding.ts` · `accuracy.ts` · `freshness.ts` · `bonus.ts` · `sources.ts`

2831

29−**Writers** (`src/writers/`): `index.ts` · `claude/index.ts` · `cursor/index.ts` · `codex/index.ts` · `github-copilot/index.ts` · `staging.ts` · `manifest.ts` · `backup.ts`

32+**Writers** (`src/writers/`): `index.ts` · `claude/index.ts` · `cursor/index.ts` · `codex/index.ts` · `opencode/index.ts` · `github-copilot/index.ts` · `refresh.ts` · `staging.ts` · `backup.ts` · `manifest.ts` · `pre-commit-block.ts`

3033

31−**Lib** (`src/lib/`): `hooks.ts` · `state.ts` · `resolve-caliber.ts` · `builtin-skills.ts` · `sanitize.ts` · `git-diff.ts`

34+**Scanner** (`src/scanner/`): `index.ts` — detects local `.mcp.json`, `.cursor/mcp.json` MCP servers, rules, and skills across platforms

3235

33−**Other**: `src/constants.ts` · `src/test/setup.ts` · `github-action/` · `packages/mcp-server/` · `packages/shared/` · `apps/`

36+**Lib** (`src/lib/`): `hooks.ts` · `learning-hooks.ts` · `state.ts` · `resolve-caliber.ts` · `builtin-skills.ts` · `sanitize.ts` · `notifications.ts` · `git-diff.ts` · `lock.ts` · `debug-report.ts` · `config-discovery.ts` · `terminal.ts`

3437

38+**Utils** (`src/utils/`): `parallel-tasks.ts` · `spinner-messages.ts` · `editor.ts` · `review.ts` · `prompt.ts` · `version-check.ts` · `dependencies.ts` · `waiting-content.ts` · `waiting-cards.json`

39+

40+**Telemetry** (`src/telemetry/`): `index.ts` · `config.ts` · `events.ts` · **Learner** (`src/learner/`): `writer.ts` · `storage.ts` · `attribution.ts` · `roi.ts` · `utils.ts` · `stdin.ts`

41+

42+**Other**: `github-action/action.yml` · `github-action/index.js` · `assets/video/` (Remotion) · `scripts/` · `docs/FLOW.md` · `src/constants.ts` · `src/test/setup.ts` · `CHANGELOG.md` · `TODOS.md`

43+

44+**Workspaces**: `packages/shared/` · `packages/mcp-server/` (MCP server) · `apps/web/` · `apps/api/`

45+

46+@./CONTRIBUTING.md

47+

3548 ## Conventions

3649

37−- ESM with `.js` extensions · Tests in `__tests__/` dirs · `unknown` over `any`

38−- Conventional commits (`feat:`, `fix:`, `refactor:`) · Dev: `next`, Stable: `master`

39−- Providers implement `LLMProvider` from `src/llm/types.ts`

40−- Writers in `src/writers/claude/index.ts` · `src/writers/codex/index.ts` return `string[]`

41−- Scoring checks in `src/scoring/checks/` return `Check[]`

42−- No hardcoded mappings — LLM-driven via `src/ai/detect.ts`

43−- `collectFingerprint()` async with LLM enrichment (`src/fingerprint/index.ts`)

44−- MCP server workspace: `packages/mcp-server/`

50+- ESM with `.js` import extensions · Tests in `__tests__/` dirs · Setup: `src/test/setup.ts`

51+- `unknown` over `any` · Conventional commits (`feat:`, `fix:`, `refactor:`)

52+- Dev: `next` branch · Stable: `master` · Node >= 20

53+- Config: `~/.caliber/config.json` (mode `0600`) · Constants: `src/constants.ts`

54+- `package.json` scripts: `build`, `dev`, `test`, `lint`, `lint:fix`, `format`, `format:check`

4555

56+## Key Patterns

57+

58+- **Providers**: implement `LLMProvider` from `src/llm/types.ts` (`call()`, `stream()`)

59+- **Writers**: `src/writers/claude/index.ts` · `src/writers/cursor/index.ts` · `src/writers/codex/index.ts` · `src/writers/opencode/index.ts` each return `string[]`

60+- **Scoring**: checks in `src/scoring/checks/` return `Check[]`, constants in `src/scoring/constants.ts`

61+- **No hardcoded mappings**: detection is LLM-driven via `src/ai/detect.ts`

62+- `collectFingerprint()` in `src/fingerprint/index.ts` — async with LLM enrichment

63+- `resolveCaliber()` from `src/lib/resolve-caliber.ts` for CLI binary resolution

64+- Cursor provider: `agent --print --trust --workspace /tmp` in `src/llm/cursor-acp.ts`

65+- `getFastModel()` in `src/llm/config.ts`: `ANTHROPIC_SMALL_FAST_MODEL` scoped to anthropic/vertex

66+- `validateModel()` skips seat-based providers (`isSeatBased()` in `src/llm/types.ts`)

67+- Scoring: deterministic, no LLM · Backups via `src/writers/backup.ts`

68+- Scanner: `src/scanner/index.ts` detects local MCP servers, rules, skills for state comparison

69+- MCP server workspace: `packages/mcp-server/` · Config discovery: `src/lib/config-discovery.ts`

70+- Builtin skills (`src/lib/builtin-skills.ts`): `find-skills`, `save-learning`, `setup-caliber`

71+

4672

4773 ## Before Committing

4874

@@ −56 +82 @@

5682 - If **hook-active**: the hook handles sync automatically — just commit normally. Tell the user: "Caliber will sync your agent configs automatically via the pre-commit hook."

5783 - If **no-hook**: run Caliber manually before committing:

5884 1. Tell the user: "Caliber: Syncing agent configs with your latest changes..."

59− 2. Run: `caliber refresh && git add CLAUDE.md .claude/ .cursor/ .cursorrules .github/copilot-instructions.md .github/instructions/ AGENTS.md CALIBER_LEARNINGS.md 2>/dev/null`

85+ 2. Run: `caliber refresh && git add CLAUDE.md .claude/ .cursor/ .cursorrules .github/copilot-instructions.md .github/instructions/ AGENTS.md CALIBER_LEARNINGS.md .agents/ .opencode/ 2>/dev/null`

6086 3. After it completes, briefly tell the user what Caliber updated. Then proceed with the commit.

6187

6288 If `caliber` is not found, tell the user: "This project uses Caliber for agent config sync. Run /setup-caliber to get set up."

@@ −68 +94 @@

6894 Read `CALIBER_LEARNINGS.md` for patterns and anti-patterns learned from previous sessions.

6995 These are auto-extracted from real tool usage — treat them as project-specific rules.

7096

97+

98+

99+## Context Sync

100+

101+This project uses [Caliber](https://github.com/caliber-ai-org/ai-setup) to keep AI agent configs in sync across Claude Code, Cursor, Copilot, and Codex.

102+Configs update automatically before each commit via `caliber refresh`.

103+If the pre-commit hook is not set up, run `/setup-caliber` to configure everything automatically.

104+

71105

Two files, one repository

Built for the same person as RuleStack