@rely-ai/caliber 1.36.1 → 1.37.0-dev.1774893051

package/README.md

@@ -1,3 +1,7 @@
+# Caliber
+
+**Hand-written `CLAUDE.md` files go stale the moment you refactor.** Your AI agent hallucinates paths that no longer exist, misses new dependencies, and gives advice based on yesterday's architecture. Caliber generates and maintains your AI context files (`CLAUDE.md`, `.cursor/rules/`, `AGENTS.md`, `copilot-instructions.md`) so they stay accurate as your code evolves — and keeps every agent on your team in sync, whether they use Claude Code, Cursor, Codex, OpenCode, or GitHub Copilot.
+
 <p align="center">
   <img src="assets/demo-header.gif" alt="Caliber product demo" width="900">
 </p>
@@ -10,30 +14,16 @@
   <img src="https://img.shields.io/badge/Claude_Code-supported-blue" alt="Claude Code">
   <img src="https://img.shields.io/badge/Cursor-supported-blue" alt="Cursor">
   <img src="https://img.shields.io/badge/Codex-supported-blue" alt="Codex">
+  <img src="https://img.shields.io/badge/OpenCode-supported-blue" alt="OpenCode">
+  <img src="https://img.shields.io/badge/GitHub_Copilot-supported-blue" alt="GitHub Copilot">
 </p>
 
----
-
-### Try it — zero install, zero commitment
-
-```bash
-npx @rely-ai/caliber score
-```
-
-Score your AI agent config in 3 seconds. No API key. No changes to your code. Just a score.
-
-> **Your code stays on your machine.** Scoring is 100% local — no LLM calls, no code sent anywhere. Generation uses your own AI subscription (Claude Code, Cursor) or your own API key (Anthropic, OpenAI, Vertex AI). Caliber never sees your code.
-
----
-
-Caliber scores, generates, and keeps your AI agent configs in sync with your codebase. It fingerprints your project — languages, frameworks, dependencies, architecture — and produces tailored configs for **Claude Code**, **Cursor**, and **OpenAI Codex**. When your code evolves, Caliber detects the drift and updates your configs to match.
-
 ## Before / After
 
 Most repos start with a hand-written `CLAUDE.md` and nothing else. Here's what Caliber finds — and fixes:
 
 ```
-  Before                                    After caliber init
+  Before                                    After /setup-caliber
   ──────────────────────────────            ──────────────────────────────
 
   Agent Config Score    35 / 100            Agent Config Score    94 / 100
@@ -53,6 +43,24 @@ Scoring is deterministic — no LLM, no API calls. It cross-references your conf
 caliber score --compare main    # See how your branch changed the score
 ```
 
+## Get Started
+
+Requires **Node.js >= 20**.
+
+```bash
+npx @rely-ai/caliber bootstrap
+```
+
+Then, in your next Claude Code or Cursor chat session, type:
+
+> **/setup-caliber**
+
+Your agent detects your stack, generates tailored configs for every platform your team uses, sets up pre-commit hooks, and enables continuous sync — all from inside your normal workflow.
+
+**Don't use Claude Code or Cursor?** Run `caliber init` instead — it's the same setup as a CLI wizard. Works with any LLM provider: bring your own Anthropic, OpenAI, or Vertex AI key.
+
+> **Your code stays on your machine.** Bootstrap is 100% local — no LLM calls, no code sent anywhere. Generation uses your own AI subscription or API key. Caliber never sees your code.
+
 ## Audits first, writes second
 
 Caliber never overwrites your existing configs without asking. The workflow mirrors code review:
@@ -67,26 +75,28 @@ If your existing config scores **95+**, Caliber skips full regeneration and appl
 
 ## How It Works
 
-Caliber is not a one-time setup tool. It's a loop:
+Bootstrap gives your agent the `/setup-caliber` skill. Your agent analyzes your project — languages, frameworks, dependencies, architecture — generates configs, and installs hooks. From there, it's a loop:
 
 ```
-          caliber score
+  npx @rely-ai/caliber bootstrap       ← one-time, 2 seconds
               │
               ▼
-  ┌──── caliber init ◄────────────────┐
-  │     (generate / fix)              │
-  │           │                       │
-  │           ▼                       │
-  │     your code evolves             │
-  │     (new deps, renamed files,     │
-  │      changed architecture)        │
-  │           │                       │
-  │           ▼                       │
-  └──► caliber refresh ──────────────►┘
-       (detect drift, update configs)
+  agent runs /setup-caliber             ← agent handles everything
+              │
+              ▼
+  ┌──── configs generated ◄────────────┐
+  │           │                        │
+  │           ▼                        │
+  │     your code evolves              │
+  │     (new deps, renamed files,      │
+  │      changed architecture)         │
+  │           │                        │
+  │           ▼                        │
+  └──► caliber refresh ──────────────►─┘
+       (auto, on every commit)
 ```
 
-Auto-refresh hooks run this loop automatically — on every commit or at the end of each AI coding session.
+Pre-commit hooks run the refresh loop automatically. New team members get nudged to bootstrap on their first session.
 
 ### What It Generates
 
@@ -106,6 +116,13 @@ Auto-refresh hooks run this loop automatically — on every commit or at the end
 - `AGENTS.md` — Project context for Codex
 - `.agents/skills/*/SKILL.md` — Skills for Codex
 
+**OpenCode**
+- `AGENTS.md` — Project context (shared with Codex when both are targeted)
+- `.opencode/skills/*/SKILL.md` — Skills for OpenCode
+
+**GitHub Copilot**
+- `.github/copilot-instructions.md` — Project context for Copilot
+
 ## Key Features
 
 <details>
@@ -118,13 +135,15 @@ TypeScript, Python, Go, Rust, Java, Ruby, Terraform, and more. Language and fram
 <details>
 <summary><strong>Any AI Tool</strong></summary>
 
-Target a single platform or all three at once:
+`caliber bootstrap` auto-detects which agents you have installed. For manual control:
 ```bash
 caliber init --agent claude        # Claude Code only
 caliber init --agent cursor        # Cursor only
 caliber init --agent codex         # Codex only
-caliber init --agent all           # All three
-caliber init --agent claude,cursor # Comma-separated
+caliber init --agent opencode        # OpenCode only
+caliber init --agent github-copilot  # GitHub Copilot only
+caliber init --agent all             # All platforms
+caliber init --agent claude,cursor   # Comma-separated
 ```
 
 </details>
@@ -197,12 +216,20 @@ The `refresh` command analyzes your git diff (committed, staged, and unstaged ch
 
 </details>
 
+<details>
+<summary><strong>Team Onboarding</strong></summary>
+
+When Caliber is set up in a repo, it automatically nudges new team members to configure it on their machine. A lightweight session hook checks whether the pre-commit hook is installed and prompts setup if not — no manual coordination needed.
+
+</details>
+
 <details>
 <summary><strong>Fully Reversible</strong></summary>
 
 - **Automatic backups** — originals saved to `.caliber/backups/` before every write
 - **Score regression guard** — if a regeneration produces a lower score, changes are auto-reverted
 - **Full undo** — `caliber undo` restores everything to its previous state
+- **Clean uninstall** — `caliber uninstall` removes everything Caliber added (hooks, generated sections, skills, learnings) while preserving your own content
 - **Dry run** — preview changes with `--dry-run` before applying
 
 </details>
@@ -211,9 +238,10 @@ The `refresh` command analyzes your git diff (committed, staged, and unstaged ch
 
 | Command | Description |
 |---|---|
+| `caliber bootstrap` | Install agent skills — the fastest way to get started |
+| `caliber init` | Full setup wizard — analyze, generate, review, install hooks |
 | `caliber score` | Score config quality (deterministic, no LLM) |
 | `caliber score --compare <ref>` | Compare current score against a git ref |
-| `caliber init` | Full setup wizard — analyze, generate, review, install hooks |
 | `caliber regenerate` | Re-analyze and regenerate configs (aliases: `regen`, `re`) |
 | `caliber refresh` | Update docs based on recent code changes |
 | `caliber skills` | Discover and install community skills |
@@ -221,6 +249,7 @@ The `refresh` command analyzes your git diff (committed, staged, and unstaged ch
 | `caliber hooks` | Manage auto-refresh hooks |
 | `caliber config` | Configure LLM provider, API key, and model |
 | `caliber status` | Show current setup status |
+| `caliber uninstall` | Remove all Caliber resources from a project |
 | `caliber undo` | Revert all changes made by Caliber |
 
 ## FAQ
@@ -235,9 +264,16 @@ No. Caliber shows you a diff of every proposed change. You accept, refine, or de
 <details>
 <summary><strong>Does it need an API key?</strong></summary>
 
-**Scoring:** No. `caliber score` runs 100% locally with no LLM.
+**Bootstrap & scoring:** No. Both run 100% locally with no LLM.
 
-**Generation:** Uses your existing Claude Code or Cursor subscription (no API key needed), or bring your own key for Anthropic, OpenAI, or Vertex AI.
+**Generation** (via `/setup-caliber` or `caliber init`): Uses your existing Claude Code or Cursor subscription (no API key needed), or bring your own key for Anthropic, OpenAI, or Vertex AI.
+
+</details>
+
+<details>
+<summary><strong>What's the difference between bootstrap and init?</strong></summary>
+
+`caliber bootstrap` installs agent skills in 2 seconds — your agent then runs `/setup-caliber` to handle the rest from inside your session. `caliber init` is the full interactive wizard for users who prefer a CLI-driven setup. Both end up in the same place.
 
 </details>
 
@@ -258,24 +294,10 @@ Yes. Run `caliber init` from any directory. `caliber refresh` can update configs
 <details>
 <summary><strong>Does it send my code anywhere?</strong></summary>
 
-Scoring is fully local. Generation sends your project fingerprint (not source code) to whatever LLM provider you configure — the same provider your AI editor already uses. Anonymous usage analytics (no code, no file contents) can be disabled via `caliber config`.
+Scoring is fully local. Generation sends a project summary (languages, structure, dependencies — not source code) to whatever LLM provider you configure — the same provider your AI editor already uses. Anonymous usage analytics (no code, no file contents) can be disabled via `caliber config`.
 
 </details>
 
-## Add a Caliber badge to your repo
-
-After scoring your project, add a badge to your README:
-
-![Caliber Score](https://img.shields.io/badge/caliber-94%2F100-brightgreen)
-
-Copy this markdown and replace `94` with your actual score:
-
-```
-![Caliber Score](https://img.shields.io/badge/caliber-SCORE%2F100-COLOR)
-```
-
-Color guide: `brightgreen` (90+), `green` (70-89), `yellow` (40-69), `red` (<40).
-
 ## LLM Providers
 
 No API key? No problem. Caliber works with your existing AI tool subscription:
@@ -285,9 +307,9 @@ No API key? No problem. Caliber works with your existing AI tool subscription:
 | **Claude Code** (your seat) | `caliber config` → Claude Code | Inherited from Claude Code |
 | **Cursor** (your seat) | `caliber config` → Cursor | Inherited from Cursor |
 | **Anthropic** | `export ANTHROPIC_API_KEY=sk-ant-...` | `claude-sonnet-4-6` |
-| **OpenAI** | `export OPENAI_API_KEY=sk-...` | `gpt-5.4-mini` |
+| **OpenAI** | `export OPENAI_API_KEY=sk-...` | `gpt-4.1` |
 | **Vertex AI** | `export VERTEX_PROJECT_ID=my-project` | `claude-sonnet-4-6` |
-| **Custom endpoint** | `OPENAI_API_KEY` + `OPENAI_BASE_URL` | `gpt-5.4-mini` |
+| **Custom endpoint** | `OPENAI_API_KEY` + `OPENAI_BASE_URL` | `gpt-4.1` |
 
 Override the model for any provider: `export CALIBER_MODEL=<model-name>` or use `caliber config`.
 
@@ -333,11 +355,6 @@ export GOOGLE_APPLICATION_CREDENTIALS=/path/to/service-account.json
 
 </details>
 
-## Requirements
-
-- **Node.js** >= 20
-- **One LLM provider:** your **Claude Code** or **Cursor** subscription (no API key), or an API key for Anthropic / OpenAI / Vertex AI
-
 ## Contributing
 
 See [CONTRIBUTING.md](./CONTRIBUTING.md) for detailed guidelines.
@@ -353,6 +370,20 @@ npm run build    # Compile
 
 Uses [conventional commits](https://www.conventionalcommits.org/) — `feat:` for features, `fix:` for bug fixes.
 
+## Add a Caliber badge to your repo
+
+After scoring your project, add a badge to your README:
+
+![Caliber Score](https://img.shields.io/badge/caliber-94%2F100-brightgreen)
+
+Copy this markdown and replace `94` with your actual score:
+
+```
+![Caliber Score](https://img.shields.io/badge/caliber-SCORE%2F100-COLOR)
+```
+
+Color guide: `brightgreen` (90+), `green` (70-89), `yellow` (40-69), `red` (<40).
+
 ## License
 
 MIT

package/dist/bin.js

@@ -35,9 +35,9 @@ function getMaxPromptTokens() {
   return Math.max(MIN_PROMPT_TOKENS, Math.min(budget, MAX_PROMPT_TOKENS_CAP));
 }
 function loadConfig() {
-  const fileConfig = readConfigFile();
-  if (fileConfig) return fileConfig;
-  return resolveFromEnv();
+  const envConfig = resolveFromEnv();
+  if (envConfig) return envConfig;
+  return readConfigFile();
 }
 function resolveFromEnv() {
   if (process.env.ANTHROPIC_API_KEY) {
@@ -130,7 +130,7 @@ var init_config = __esm({
     DEFAULT_MODELS = {
       anthropic: "claude-sonnet-4-6",
       vertex: "claude-sonnet-4-6",
-      openai: "gpt-5.4-mini",
+      openai: "gpt-4.1",
       cursor: "sonnet-4.6",
       "claude-cli": "default"
     };
@@ -139,7 +139,8 @@ var init_config = __esm({
       "claude-opus-4-6": 2e5,
       "claude-haiku-4-5-20251001": 2e5,
       "claude-sonnet-4-5-20250514": 2e5,
-      "gpt-5.4-mini": 1e6,
+      "gpt-4.1": 1e6,
+      "gpt-4.1-mini": 1e6,
       "gpt-4o": 128e3,
       "gpt-4o-mini": 128e3,
       "sonnet-4.6": 2e5
@@ -151,7 +152,7 @@ var init_config = __esm({
     DEFAULT_FAST_MODELS = {
       anthropic: "claude-haiku-4-5-20251001",
       vertex: "claude-haiku-4-5-20251001",
-      openai: "gpt-5.4-mini",
+      openai: "gpt-4.1-mini",
       cursor: "gpt-5.3-codex-fast"
     };
   }
@@ -2835,7 +2836,13 @@ var KNOWN_MODELS = {
     "claude-opus-4-6@20250605",
     "claude-opus-4-1-20250620"
   ],
-  openai: ["gpt-5.4-mini", "gpt-4o", "gpt-4o-mini", "o3-mini"],
+  openai: [
+    "gpt-4.1",
+    "gpt-4.1-mini",
+    "gpt-4o",
+    "gpt-4o-mini",
+    "o3-mini"
+  ],
   cursor: ["auto", "composer-1.5"],
   "claude-cli": []
 };
@@ -2843,8 +2850,7 @@ function isModelNotAvailableError(error) {
   const msg = error.message.toLowerCase();
   const status = error.status;
   if (status === 404 && msg.includes("model")) return true;
-  if (msg.includes("model") && (msg.includes("not found") || msg.includes("not_found")))
-    return true;
+  if (msg.includes("model") && (msg.includes("not found") || msg.includes("not_found"))) return true;
   if (msg.includes("model") && msg.includes("not available")) return true;
   if (msg.includes("model") && msg.includes("does not exist")) return true;
   if (msg.includes("publisher model")) return true;
@@ -2870,18 +2876,12 @@ function filterRelevantModels(models, provider) {
 async function handleModelNotAvailable(failedModel, provider, config) {
   if (!process.stdin.isTTY) {
     console.error(
-      chalk.red(
-        `Model "${failedModel}" is not available. Run \`${resolveCaliber()} config\` to select a different model.`
-      )
+      chalk.red(`Model "${failedModel}" is not available. Run \`${resolveCaliber()} config\` to select a different model.`)
     );
     return null;
   }
-  console.log(
-    chalk.yellow(
-      `
-\u26A0  Model "${failedModel}" is not available on your ${config.provider} deployment.`
-    )
-  );
+  console.log(chalk.yellow(`
+\u26A0  Model "${failedModel}" is not available on your ${config.provider} deployment.`));
   let models = [];
   if (provider.listModels) {
     try {
@@ -2895,11 +2895,7 @@ async function handleModelNotAvailable(failedModel, provider, config) {
   }
   models = models.filter((m) => m !== failedModel);
   if (models.length === 0) {
-    console.log(
-      chalk.red(
-        `  No alternative models found. Run \`${resolveCaliber()} config\` to configure manually.`
-      )
-    );
+    console.log(chalk.red(`  No alternative models found. Run \`${resolveCaliber()} config\` to configure manually.`));
     return null;
   }
   console.log("");
@@ -3105,6 +3101,7 @@ var SKILL_FORMAT_RULES = `All skills follow the OpenSkills standard (agentskills
 
 Skill field requirements:
 - "name": kebab-case (lowercase letters, numbers, hyphens only). Becomes the directory name.
+- "name" MUST NOT be any of these reserved names (they are managed by Caliber automatically): "setup-caliber", "find-skills", "save-learning". Do NOT generate skills with these names.
 - "description": MUST include WHAT it does + WHEN to use it with specific trigger phrases. Example: "Manages database migrations. Use when user says 'run migration', 'create migration', 'db schema change', or modifies files in db/migrations/."
 - "content": markdown body only \u2014 do NOT include YAML frontmatter, it is generated from name+description.
 
@@ -3148,7 +3145,7 @@ PRIORITY WHEN CONSTRAINTS CONFLICT: Grounding and reference density matter more
 Note: Permissions, hooks, freshness tracking, and OpenSkills frontmatter are scored automatically by caliber \u2014 do not optimize for them.
 README.md is provided for context only \u2014 do NOT include a readmeMd field in your output.`;
 var OUTPUT_SIZE_CONSTRAINTS = `OUTPUT SIZE CONSTRAINTS \u2014 these are critical:
-- CLAUDE.md / AGENTS.md: MUST be under 150 lines for maximum score. Aim for 100-140 lines. Be concise \u2014 commands, architecture overview, and key conventions. Use bullet points and tables, not prose.
+- CLAUDE.md / AGENTS.md: MUST be under 400 lines for maximum score. Aim for 200-350 lines. Be thorough \u2014 commands, architecture overview, key conventions, data flow, and important patterns. Use bullet points and tables, not prose.
 
 Pack project references densely in architecture sections \u2014 use inline paths, not prose paragraphs:
 GOOD: **Entry**: \`src/bin.ts\` \u2192 \`src/cli.ts\` \xB7 **LLM** (\`src/llm/\`): \`anthropic.ts\` \xB7 \`vertex.ts\` \xB7 \`openai-compat.ts\`
@@ -3286,7 +3283,7 @@ Structure:
 5. "## Common Issues" (required) \u2014 specific error messages and their fixes. Not "check your config" but "If you see 'Connection refused on port 5432': 1. Verify postgres is running: docker ps | grep postgres 2. Check .env has correct DATABASE_URL"
 
 Rules:
-- Max 150 lines. Focus on actionable instructions, not documentation prose.
+- Max 400 lines. Focus on actionable instructions, not documentation prose.
 - Study existing code in the project context to extract the real patterns being used. A skill for "create API route" should show the exact file structure, imports, error handling, and naming that existing routes use.
 - Be specific and actionable. GOOD: "Run \`pnpm test -- --filter=api\` to verify". BAD: "Validate the data before proceeding."
 - Never use ambiguous language. Instead of "handle errors properly", write "Wrap the DB call in try/catch. On failure, return { error: string, code: number } matching the ErrorResponse type in \`src/types.ts\`."
@@ -3344,7 +3341,7 @@ Rules:
 - Update the "fileDescriptions" to reflect any changes you make.
 
 Quality constraints \u2014 your changes are scored, so do not break these:
-- CLAUDE.md / AGENTS.md: MUST stay under 150 lines. If adding content, remove less important lines to stay within budget. Do not refuse the user's request \u2014 make the change and trim elsewhere.
+- CLAUDE.md / AGENTS.md: MUST stay under 400 lines. If adding content, remove less important lines to stay within budget. Do not refuse the user's request \u2014 make the change and trim elsewhere.
 - Avoid vague instructions ("follow best practices", "write clean code", "ensure quality").
 - Do NOT add directory tree listings in code blocks.
 - Do NOT remove existing code blocks \u2014 they contribute to the executable content score.
@@ -3368,7 +3365,7 @@ CONSERVATIVE UPDATE means:
 - NEVER replace specific paths/commands with generic prose
 
 Quality constraints (the output is scored deterministically):
-- CLAUDE.md / AGENTS.md: MUST stay under 150 lines. If the diff adds content, trim the least important lines elsewhere.
+- CLAUDE.md / AGENTS.md: MUST stay under 400 lines. If the diff adds content, trim the least important lines elsewhere.
 - Keep 3+ code blocks with executable commands \u2014 do not remove code blocks
 - Every file path, command, and identifier must be in backticks
 - ONLY reference file paths that exist in the provided file tree \u2014 do NOT invent paths
@@ -3381,6 +3378,7 @@ Cross-agent sync:
 
 Managed content:
 - Keep managed blocks (<!-- caliber:managed --> ... <!-- /caliber:managed -->) intact
+- Keep context sync blocks (<!-- caliber:managed:sync --> ... <!-- /caliber:managed:sync -->) intact
 - Do NOT modify CALIBER_LEARNINGS.md \u2014 it is managed separately
 - Preserve any references to CALIBER_LEARNINGS.md in CLAUDE.md
 
@@ -3396,9 +3394,12 @@ Return a JSON object with this exact shape:
     "copilotInstructionFiles": [{"filename": "name.instructions.md", "content": "..."}] or null
   },
   "changesSummary": "<1-2 sentence summary of what was updated and why>",
+  "fileChanges": [{"file": "CLAUDE.md", "description": "added new API routes, updated build commands"}],
   "docsUpdated": ["CLAUDE.md", "README.md"]
 }
 
+The "fileChanges" array MUST include one entry per file that was updated (non-null in updatedDocs). Each entry describes what specifically changed in that file \u2014 be concrete (e.g. "added auth middleware section" not "updated docs").
+
 Respond with ONLY the JSON object, no markdown fences or extra text.`;
 var LEARN_SYSTEM_PROMPT = `You are an expert developer experience engineer. You analyze raw tool call events from AI coding sessions to extract reusable operational lessons that will help future LLM sessions work more effectively in this project.
 
@@ -5258,7 +5259,7 @@ import fs12 from "fs";
 import path12 from "path";
 function writeClaudeConfig(config) {
   const written = [];
-  fs12.writeFileSync("CLAUDE.md", appendLearningsBlock(appendPreCommitBlock(config.claudeMd)));
+  fs12.writeFileSync("CLAUDE.md", appendSyncBlock(appendLearningsBlock(appendPreCommitBlock(config.claudeMd))));
   written.push("CLAUDE.md");
   if (config.skills?.length) {
     for (const skill of config.skills) {
@@ -5304,7 +5305,8 @@ function writeCursorConfig(config) {
   }
   const preCommitRule = getCursorPreCommitRule();
   const learningsRule = getCursorLearningsRule();
-  const allRules = [...config.rules || [], preCommitRule, learningsRule];
+  const syncRule = getCursorSyncRule();
+  const allRules = [...config.rules || [], preCommitRule, learningsRule, syncRule];
   const rulesDir = path13.join(".cursor", "rules");
   if (!fs13.existsSync(rulesDir)) fs13.mkdirSync(rulesDir, { recursive: true });
   for (const rule of allRules) {
@@ -5387,7 +5389,7 @@ function writeGithubCopilotConfig(config) {
     fs15.mkdirSync(".github", { recursive: true });
     fs15.writeFileSync(
       path15.join(".github", "copilot-instructions.md"),
-      appendLearningsBlock(appendPreCommitBlock(config.instructions, "copilot"))
+      appendSyncBlock(appendLearningsBlock(appendPreCommitBlock(config.instructions, "copilot")))
     );
     written.push(".github/copilot-instructions.md");
   }
@@ -6120,11 +6122,11 @@ var POINTS_LEARNED_CONTENT = 2;
 var POINTS_SOURCES_CONFIGURED = 3;
 var POINTS_SOURCES_REFERENCED = 3;
 var TOKEN_BUDGET_THRESHOLDS = [
-  { maxTokens: 2e3, points: 6 },
-  { maxTokens: 3500, points: 5 },
-  { maxTokens: 5e3, points: 4 },
-  { maxTokens: 8e3, points: 2 },
-  { maxTokens: 12e3, points: 1 }
+  { maxTokens: 5e3, points: 6 },
+  { maxTokens: 8e3, points: 5 },
+  { maxTokens: 12e3, points: 4 },
+  { maxTokens: 16e3, points: 2 },
+  { maxTokens: 24e3, points: 1 }
 ];
 var CODE_BLOCK_THRESHOLDS = [
   { minBlocks: 3, points: 8 },
@@ -6292,17 +6294,14 @@ function checkExistence(dir) {
   const opencodeSkills = countFiles(join3(dir, ".opencode", "skills"), /SKILL\.md$/);
   const skillCount = claudeSkills.length + codexSkills.length + opencodeSkills.length;
   const skillBase = skillCount >= 1 ? POINTS_SKILLS_EXIST : 0;
-  const skillBonus = Math.min(
-    (skillCount - 1) * POINTS_SKILLS_BONUS_PER_EXTRA,
-    POINTS_SKILLS_BONUS_CAP
-  );
+  const skillBonus = Math.min((skillCount - 1) * POINTS_SKILLS_BONUS_PER_EXTRA, POINTS_SKILLS_BONUS_CAP);
   const skillPoints = skillCount >= 1 ? skillBase + Math.max(0, skillBonus) : 0;
   const maxSkillPoints = POINTS_SKILLS_EXIST + POINTS_SKILLS_BONUS_CAP;
   checks.push({
     id: "skills_exist",
     name: "Skills configured",
     category: "existence",
-    maxPoints: skillCount >= 1 ? maxSkillPoints : 0,
+    maxPoints: maxSkillPoints,
     earnedPoints: Math.min(skillPoints, maxSkillPoints),
     passed: skillCount >= 1,
     detail: skillCount === 0 ? "No skills found" : `${skillCount} skill${skillCount === 1 ? "" : "s"} found`,
@@ -6335,7 +6334,7 @@ function checkExistence(dir) {
     id: "mcp_servers",
     name: "MCP servers configured",
     category: "existence",
-    maxPoints: mcp.count >= 1 ? POINTS_MCP_SERVERS : 0,
+    maxPoints: POINTS_MCP_SERVERS,
     earnedPoints: mcp.count >= 1 ? POINTS_MCP_SERVERS : 0,
     passed: mcp.count >= 1,
     detail: mcp.count > 0 ? `${mcp.count} server${mcp.count === 1 ? "" : "s"} in ${mcp.sources.join(", ")}` : "No MCP servers configured",
@@ -6870,11 +6869,7 @@ init_resolve_caliber();
 init_pre_commit_block();
 function hasPreCommitHook(dir) {
   try {
-    const gitDir = execSync12("git rev-parse --git-dir", {
-      cwd: dir,
-      encoding: "utf-8",
-      stdio: ["pipe", "pipe", "pipe"]
-    }).trim();
+    const gitDir = execSync12("git rev-parse --git-dir", { cwd: dir, encoding: "utf-8", stdio: ["pipe", "pipe", "pipe"] }).trim();
     const hookPath = join7(gitDir, "hooks", "pre-commit");
     const content = readFileOrNull(hookPath);
     return content ? content.includes("caliber") : false;
@@ -6965,9 +6960,9 @@ function checkBonus(dir) {
     id: "open_skills_format",
     name: "Skills use OpenSkills format",
     category: "bonus",
-    maxPoints: totalSkillFiles > 0 ? POINTS_OPEN_SKILLS_FORMAT : 0,
+    maxPoints: POINTS_OPEN_SKILLS_FORMAT,
     earnedPoints: allOpenSkills ? POINTS_OPEN_SKILLS_FORMAT : 0,
-    passed: allOpenSkills || totalSkillFiles === 0,
+    passed: allOpenSkills,
     detail: totalSkillFiles === 0 ? "No skills to check" : allOpenSkills ? `All ${totalSkillFiles} skill${totalSkillFiles === 1 ? "" : "s"} use SKILL.md with frontmatter` : `${openSkillsCount}/${totalSkillFiles} use OpenSkills format`,
     suggestion: totalSkillFiles > 0 && !allOpenSkills ? "Migrate skills to .claude/skills/{name}/SKILL.md with YAML frontmatter" : void 0,
     fix: totalSkillFiles > 0 && !allOpenSkills ? {
@@ -9773,22 +9768,11 @@ async function initCommand(options) {
     );
     console.log(chalk14.dim("  Keep your AI agent configs in sync \u2014 automatically."));
     console.log(chalk14.dim("  Works across Claude Code, Cursor, Codex, and GitHub Copilot.\n"));
-    console.log(title.bold("  What this does:\n"));
-    console.log(
-      chalk14.dim("  Caliber reads your project structure (file tree, package.json, etc.)")
-    );
-    console.log(chalk14.dim("  and generates agent config files (CLAUDE.md, .cursor/rules/, etc.)."));
-    console.log(chalk14.dim("  You review all changes before anything is written to disk.\n"));
-    console.log(title.bold("  Steps:\n"));
-    console.log(
-      chalk14.dim("  1. Connect    Pick your LLM provider (or use your existing subscription)")
-    );
-    console.log(
-      chalk14.dim("  2. Build      Scan project, generate configs, install pre-commit sync")
-    );
-    console.log(
-      chalk14.dim("  3. Review     See exactly what changed \u2014 accept, refine, or decline\n")
-    );
+    console.log(title.bold("  How it works:\n"));
+    console.log(chalk14.dim("  1. Connect    Link your LLM provider and select your agents"));
+    console.log(chalk14.dim("  2. Setup      Detect stack, install sync hooks & skills"));
+    console.log(chalk14.dim("  3. Generate   Audit existing config or generate from scratch"));
+    console.log(chalk14.dim("  4. Finalize   Review changes and score your setup\n"));
   } else {
     console.log(brand.bold("\n  CALIBER") + chalk14.dim("  \u2014 setting up continuous sync\n"));
   }
@@ -9886,10 +9870,11 @@ async function initCommand(options) {
   console.log(chalk14.dim(`  Target: ${targetAgent.join(", ")}
 `));
   trackInitAgentSelected(targetAgent, agentAutoDetected);
-  console.log(title.bold("  Step 2/3 \u2014 Build\n"));
+  console.log(title.bold("  Step 2/4 \u2014 Setup\n"));
+  console.log(chalk14.dim("  Installing sync infrastructure...\n"));
   const hookResult = installPreCommitHook();
   if (hookResult.installed) {
-    console.log(`  ${chalk14.green("\u2713")} Pre-commit hook installed`);
+    console.log(`  ${chalk14.green("\u2713")} Pre-commit hook installed \u2014 configs sync on every commit`);
   } else if (hookResult.alreadyInstalled) {
     console.log(`  ${chalk14.green("\u2713")} Pre-commit hook \u2014 active`);
   }
@@ -9906,7 +9891,11 @@ async function initCommand(options) {
   }
   const skillsWritten = ensureBuiltinSkills2();
   if (skillsWritten.length > 0) {
-    console.log(`  ${chalk14.green("\u2713")} Agent skills installed`);
+    console.log(
+      `  ${chalk14.green("\u2713")} Agent skills installed \u2014 /setup-caliber, /find-skills, /save-learning`
+    );
+  } else {
+    console.log(`  ${chalk14.green("\u2713")} Agent skills \u2014 already installed`);
   }
   const hasLearnableAgent = targetAgent.includes("claude") || targetAgent.includes("cursor");
   if (hasLearnableAgent) {
@@ -9916,8 +9905,19 @@ async function initCommand(options) {
     trackInitLearnEnabled(true);
   }
   console.log("");
+  console.log(chalk14.dim("  New team members can run /setup-caliber inside their coding agent"));
+  console.log(chalk14.dim("  (Claude Code or Cursor) to get set up automatically.\n"));
   const baselineScore = computeLocalScore(process.cwd(), targetAgent);
-  log(options.verbose, `Baseline score: ${baselineScore.score}/100`);
+  console.log(chalk14.dim("  Current config score:"));
+  displayScoreSummary(baselineScore);
+  if (options.verbose) {
+    for (const c of baselineScore.checks) {
+      log(
+        options.verbose,
+        `  ${c.passed ? "\u2713" : "\u2717"} ${c.name}: ${c.earnedPoints}/${c.maxPoints}${c.suggestion ? ` \u2014 ${c.suggestion}` : ""}`
+      );
+    }
+  }
   if (report) {
     report.markStep("Baseline scoring");
     report.addSection(
@@ -9943,26 +9943,27 @@ async function initCommand(options) {
   ]);
   const passingCount = baselineScore.checks.filter((c) => c.passed).length;
   const failingCount = baselineScore.checks.filter((c) => !c.passed).length;
-  trackInitScoreComputed(baselineScore.score, passingCount, failingCount, false);
   let skipGeneration = false;
-  if (hasExistingConfig && baselineScore.score === 100 && !options.force) {
-    skipGeneration = true;
+  if (hasExistingConfig && baselineScore.score === 100) {
+    trackInitScoreComputed(baselineScore.score, passingCount, failingCount, true);
+    console.log(chalk14.bold.green("\n  Your config is already optimal.\n"));
+    skipGeneration = !options.force;
   } else if (hasExistingConfig && !options.force && !options.autoApprove) {
-    const topGains = baselineScore.checks.filter((c) => !c.passed && c.maxPoints > 0).sort((a, b) => b.maxPoints - b.earnedPoints - (a.maxPoints - a.earnedPoints)).slice(0, 3);
-    console.log(chalk14.dim(`  Config score: ${baselineScore.score}/100
-`));
-    if (topGains.length > 0) {
-      console.log(chalk14.dim("  Top improvements Caliber can make:"));
-      for (const c of topGains) {
-        const pts = c.maxPoints - c.earnedPoints;
-        console.log(
-          chalk14.dim(`    +${pts} pts`) + chalk14.white(` ${c.name}`) + (c.suggestion ? chalk14.gray(` \u2014 ${c.suggestion}`) : "")
-        );
-      }
-      console.log("");
-    }
-    const improveAnswer = await confirm2({ message: "Improve your existing configs?" });
-    skipGeneration = !improveAnswer;
+    trackInitScoreComputed(baselineScore.score, passingCount, failingCount, false);
+    console.log(
+      chalk14.dim("\n  Sync infrastructure is ready. Caliber can also audit your existing")
+    );
+    console.log(chalk14.dim("  configs and improve them using AI.\n"));
+    const auditAnswer = await promptInput("  Audit and improve your existing config? (Y/n) ");
+    skipGeneration = auditAnswer.toLowerCase() === "n";
+  } else if (!hasExistingConfig && !options.force && !options.autoApprove) {
+    trackInitScoreComputed(baselineScore.score, passingCount, failingCount, false);
+    console.log(chalk14.dim("\n  Sync infrastructure is ready. Caliber can also generate tailored"));
+    console.log(chalk14.dim("  CLAUDE.md, Cursor rules, and Codex configs for your project.\n"));
+    const generateAnswer = await promptInput("  Generate agent configs? (Y/n) ");
+    skipGeneration = generateAnswer.toLowerCase() === "n";
+  } else {
+    trackInitScoreComputed(baselineScore.score, passingCount, failingCount, false);
   }
   if (skipGeneration) {
     const {
@@ -10032,6 +10033,7 @@ async function initCommand(options) {
     );
     return;
   }
+  console.log(title.bold("\n  Step 3/4 \u2014 Generate\n"));
   const genModelInfo = fastModel ? `  Using ${displayModel} for docs, ${fastModel} for skills` : `  Using ${displayModel}`;
   console.log(chalk14.dim(genModelInfo + "\n"));
   if (report) report.markStep("Generation");
@@ -10270,7 +10272,7 @@ async function initCommand(options) {
     options.verbose,
     `Generation completed: ${elapsedMs}ms, stopReason: ${genStopReason || "end_turn"}`
   );
-  console.log(title.bold("  Step 3/3 \u2014 Done\n"));
+  console.log(title.bold("  Step 4/4 \u2014 Finalize\n"));
   const setupFiles = collectSetupFiles(generatedSetup, targetAgent);
   const staged = stageFiles(setupFiles, process.cwd());
   const totalChanges = staged.newFiles + staged.modifiedFiles;
@@ -10468,26 +10470,31 @@ ${agentRefs.join(" ")}
   console.log(chalk14.bold.green("\n  Caliber is set up!\n"));
   console.log(chalk14.bold("  What's configured:\n"));
   console.log(
-    `    ${done}  Continuous sync     ${chalk14.dim("pre-commit hook keeps all agent configs in sync")}`
+    `    ${done}  Continuous sync          ${chalk14.dim("pre-commit hook keeps all agent configs in sync")}`
+  );
+  console.log(
+    `    ${done}  Config generated         ${title(`${bin} score`)} ${chalk14.dim("for full breakdown")}`
   );
-  console.log(`    ${done}  Config generated    ${chalk14.dim(`score: ${afterScore.score}/100`)}`);
   console.log(
-    `    ${done}  Agent skills        ${chalk14.dim("/setup-caliber for new team members")}`
+    `    ${done}  Agent skills             ${chalk14.dim("/setup-caliber for new team members")}`
   );
   if (hasLearnableAgent) {
-    console.log(`    ${done}  Session learning    ${chalk14.dim("learns from your corrections")}`);
+    console.log(
+      `    ${done}  Session learning         ${chalk14.dim("agent learns from your feedback")}`
+    );
   }
   if (communitySkillsInstalled > 0) {
     console.log(
-      `    ${done}  Community skills    ${chalk14.dim(`${communitySkillsInstalled} installed for your stack`)}`
+      `    ${done}  Community skills         ${chalk14.dim(`${communitySkillsInstalled} skill${communitySkillsInstalled > 1 ? "s" : ""} installed for your stack`)}`
     );
   }
   console.log(chalk14.bold("\n  What happens next:\n"));
-  console.log(chalk14.dim("    Every commit syncs your agent configs automatically."));
-  console.log(chalk14.dim("    New team members run /setup-caliber to get set up instantly.\n"));
-  console.log(`    ${title(`${bin} score`)}        Full scoring breakdown`);
-  console.log(`    ${title(`${bin} skills`)}       Find community skills`);
-  console.log(`    ${title(`${bin} undo`)}         Revert changes`);
+  console.log(chalk14.dim("    Every commit will automatically sync your agent configs."));
+  console.log(chalk14.dim("    New team members can run /setup-caliber to get set up instantly.\n"));
+  console.log(chalk14.bold("  Explore:\n"));
+  console.log(`    ${title(`${bin} score`)}        Full scoring breakdown with improvement tips`);
+  console.log(`    ${title(`${bin} skills`)}       Find community skills for your stack`);
+  console.log(`    ${title(`${bin} undo`)}         Revert all changes from this run`);
   console.log(`    ${title(`${bin} uninstall`)}    Remove Caliber completely`);
   console.log("");
   if (options.showTokens) {
@@ -10836,27 +10843,18 @@ async function scoreCommand(options) {
   const separator = chalk18.gray("  " + "\u2500".repeat(53));
   console.log(separator);
   const bin = resolveCaliber();
-  const failing = result.checks.filter((c) => !c.passed && c.maxPoints > 0).sort((a, b) => b.maxPoints - b.earnedPoints - (a.maxPoints - a.earnedPoints));
-  if (result.score < 70 && failing.length > 0) {
-    const topFix = failing[0];
-    const pts = topFix.maxPoints - topFix.earnedPoints;
+  if (result.score < 40) {
     console.log(
-      chalk18.gray("  Biggest gain: ") + chalk18.yellow(`+${pts} pts`) + chalk18.gray(` from "${topFix.name}"`) + (topFix.suggestion ? chalk18.gray(` \u2014 ${topFix.suggestion}`) : "")
+      chalk18.gray("  Run ") + chalk18.hex("#83D1EB")(`${bin} init`) + chalk18.gray(" to generate a complete, optimized config.")
     );
+  } else if (result.score < 70) {
     console.log(
-      chalk18.gray("  Run ") + chalk18.hex("#83D1EB")(`${bin} init`) + chalk18.gray(" to auto-fix these.")
-    );
-  } else if (failing.length > 0) {
-    console.log(
-      chalk18.green("  Looking good!") + chalk18.gray(
-        ` ${failing.length} check${failing.length === 1 ? "" : "s"} can still be improved.`
-      )
+      chalk18.gray("  Run ") + chalk18.hex("#83D1EB")(`${bin} init`) + chalk18.gray(" to improve your config.")
     );
+  } else {
     console.log(
-      chalk18.gray("  Run ") + chalk18.hex("#83D1EB")(`${bin} init`) + chalk18.gray(" to improve, or ") + chalk18.hex("#83D1EB")(`${bin} regenerate`) + chalk18.gray(" to rebuild from scratch.")
+      chalk18.green("  Looking good!") + chalk18.gray(" Run ") + chalk18.hex("#83D1EB")(`${bin} regenerate`) + chalk18.gray(" to rebuild from scratch.")
     );
-  } else {
-    console.log(chalk18.green("  Perfect score! Your agent configs are fully optimized."));
   }
   console.log("");
 }
@@ -11062,6 +11060,13 @@ Changed files: ${diff.changedFiles.join(", ")}`);
     parts.push("\n[.cursorrules]");
     parts.push(existingDocs.cursorrules);
   }
+  if (existingDocs.claudeSkills?.length) {
+    for (const skill of existingDocs.claudeSkills) {
+      parts.push(`
+[.claude/skills/${skill.filename}]`);
+      parts.push(skill.content);
+    }
+  }
   if (existingDocs.cursorRules?.length) {
     for (const rule of existingDocs.cursorRules) {
       if (rule.filename.startsWith("caliber-")) continue;
@@ -11324,6 +11329,15 @@ function clearRefreshError() {
   } catch {
   }
 }
+function detectSyncedAgents(writtenFiles) {
+  const agents = [];
+  const joined = writtenFiles.join(" ");
+  if (joined.includes("CLAUDE.md") || joined.includes(".claude/")) agents.push("Claude Code");
+  if (joined.includes(".cursor/") || joined.includes(".cursorrules")) agents.push("Cursor");
+  if (joined.includes("copilot-instructions") || joined.includes(".github/instructions/")) agents.push("Copilot");
+  if (joined.includes("AGENTS.md") || joined.includes(".agents/")) agents.push("Codex");
+  return agents;
+}
 function log2(quiet, ...args) {
   if (!quiet) console.log(...args);
 }
@@ -11486,8 +11500,18 @@ async function refreshSingleRepo(repoDir, options) {
   }
   recordScore(postScore, "refresh");
   spinner?.succeed(`${prefix}Updated ${written.length} doc${written.length === 1 ? "" : "s"}`);
+  const fileChangesMap = new Map(
+    (response.fileChanges || []).map((fc) => [fc.file, fc.description])
+  );
   for (const file of written) {
-    log2(quiet, `  ${chalk19.green("\u2713")} ${file}`);
+    const desc = fileChangesMap.get(file);
+    const suffix = desc ? chalk19.dim(` \u2014 ${desc}`) : "";
+    log2(quiet, `  ${chalk19.green("\u2713")} ${file}${suffix}`);
+  }
+  const agents = detectSyncedAgents(written);
+  if (agents.length > 1) {
+    log2(quiet, chalk19.cyan(`
+  ${agents.length} agent formats in sync (${agents.join(", ")})`));
   }
   if (response.changesSummary) {
     log2(quiet, chalk19.dim(`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rely-ai/caliber",
-  "version": "1.36.1",
+  "version": "1.37.0-dev.1774893051",
   "description": "AI context infrastructure for coding agents — keeps CLAUDE.md, Cursor rules, and skills in sync as your codebase evolves",
   "type": "module",
   "bin": {