claude-launchpad 0.13.0 → 0.14.0

package/README.md

@@ -72,7 +72,7 @@ Init generates all three. Doctor checks for them. `--fix` creates any that are m
 | `claude-launchpad init` | Detect stack, generate config + hooks + permissions | Locally, free |
 | `claude-launchpad doctor --fix` | Auto-fix issues found by doctor | Locally, free |
 | `claude-launchpad eval` | Run Claude against test scenarios, score results | Via Claude CLI |
-| `claude-launchpad memory` | Optional persistent memory system | Locally |
+| `claude-launchpad memory` | Optional knowledge base that persists across sessions | Locally |
 | `/lp-enhance` (skill) | Claude reads your code and completes CLAUDE.md | Inside Claude Code |
 
 ## Doctor
@@ -193,7 +193,7 @@ Results save to `.claude/eval/` as structured markdown. Feed them back to Claude
 
 ## Memory
 
-Optional persistent memory that replaces Claude Code's built-in flat-file system. Memories decay naturally, so stale knowledge fades and relevant context stays.
+Claude maintains a personal knowledge base about your project. It learns what matters, forgets what doesn't, and gets smarter every session.
 
 ```bash
 claude-launchpad memory
@@ -201,14 +201,9 @@ claude-launchpad memory
 
 If memory is not installed, it runs interactive setup. If installed, it shows stats. Requires native deps first: `npm install better-sqlite3 sqlite-vec`.
 
-**What it does:**
-- **Smart session injection** loads the most relevant memories at session start
-- **Stop hook** extracts facts from the conversation when you finish
-- **Decay model** fades memories naturally (episodic: 30 days, semantic: 18 months, procedural: 2 years)
-- **Self-tuning retrieval** promotes memories Claude searches for, demotes ones injected but never used
-- **Project-scoped** with no cross-contamination between projects
-- **TUI dashboard** (`--dashboard`) with vim navigation, filtering, and search
-- **Cross-device sync** pushes and pulls memories between machines via private GitHub Gist
+Every session, Claude loads what it needs to know and extracts new knowledge when you finish. Stale facts fade on their own. Knowledge Claude actually uses gets reinforced. Each project has its own isolated memory, and you can sync it across machines via private GitHub Gist.
+
+Browse everything with `--dashboard` -- a terminal UI with vim navigation, filtering, and search.
 
 Data stays in `~/.agentic-memory/memory.db`. Sync requires the [GitHub CLI](https://cli.github.com/) (`gh`).
 

package/dist/cli.js

@@ -534,8 +534,10 @@ DerivedData/
 }
 
 // src/commands/init/generators/skill-enhance.ts
+var ENHANCE_SKILL_VERSION = 4;
 function generateEnhanceSkill() {
   return [
+    `<!-- lp-enhance-version: ${ENHANCE_SKILL_VERSION} -->`,
     "---",
     "name: lp-enhance",
     "description: |",
@@ -555,9 +557,11 @@ function generateEnhanceSkill() {
     "1. Read CLAUDE.md (if it exists)",
     "2. Read .claude/settings.json (hooks, permissions, MCP)",
     "3. Read .claude/rules/*.md (existing rules)",
-    "4. Scan src/ directory structure (top-level dirs, key files)",
-    "5. Read package.json / go.mod / pyproject.toml for stack detection",
-    "6. Check for monorepo indicators (workspaces, nx.json, lerna.json)",
+    "4. Read .claudeignore (if it exists)",
+    "5. Scan src/ directory structure (top-level dirs, key files)",
+    "6. Read package.json / go.mod / pyproject.toml for stack detection",
+    "7. Check for monorepo indicators (workspaces, nx.json, lerna.json)",
+    "8. Check scenarios/ directory for existing eval scenarios",
     "",
     "**Done when:** you have a mental model of the stack, architecture, and existing config.",
     "",
@@ -584,7 +588,9 @@ function generateEnhanceSkill() {
     "",
     "1. Create or update .claude/rules/ files for overflow content",
     "2. Generate path-scoped rules if the project has distinct areas (see below)",
-    "3. Verify line count is under 200",
+    "3. Review .claudeignore and print suggestions (see below)",
+    "4. Generate 2-3 custom eval scenarios in scenarios/custom/ (see below)",
+    "5. Verify line count is under 200",
     "",
     "**Rules:**",
     "- Don't remove existing content, only add or improve",
@@ -596,8 +602,9 @@ function generateEnhanceSkill() {
     "1. Run `claude-launchpad doctor` to check the score improved",
     "2. Print suggested hooks (exact JSON) for .claude/settings.json but don't modify it",
     "3. Print suggested MCP servers if external services detected (Postgres, Redis, Stripe, etc.)",
+    '4. If eval scenarios were generated, print: "Run this in your terminal (not inside Claude Code): `claude-launchpad eval --scenarios scenarios/`"',
     "",
-    "**Done when:** doctor score is equal or higher, suggestions printed.",
+    "**Done when:** doctor score is equal or higher, suggestions printed, eval scenarios created if applicable.",
     "",
     "## Path-scoped rules generation",
     "",
@@ -657,6 +664,72 @@ function generateEnhanceSkill() {
     "- If no SessionStart hook exists, suggest one that injects TASKS.md",
     "- DO NOT modify settings.json directly. Print exact JSON to add.",
     "",
+    "## .claudeignore review",
+    "",
+    "Read .claudeignore and check if the patterns make sense for the detected stack:",
+    "",
+    "**Always flag:**",
+    "- Missing node_modules/ (JS/TS projects)",
+    "- Missing __pycache__/ or .venv/ (Python projects)",
+    "- Missing target/ (Rust/Java projects)",
+    "- Missing .env / .env.* patterns",
+    "- Missing lock files (pnpm-lock.yaml, package-lock.json, yarn.lock, etc.)",
+    "- Missing coverage/ directory",
+    "- Large generated files that waste context (*.min.js, *.map, migrations/)",
+    "",
+    "**Never flag:**",
+    "- Patterns the user clearly added intentionally",
+    "- Test fixtures or seed data (might be needed for context)",
+    "",
+    "If .claudeignore is missing entirely, create one with sensible defaults for the detected stack.",
+    "If it exists but has gaps, print suggested additions. Do NOT modify it directly.",
+    "",
+    "## Eval scenario generation",
+    "",
+    "After improving CLAUDE.md, generate 2-3 custom eval scenarios that test whether Claude follows the project's specific rules. Write them as YAML files in scenarios/ at the project root.",
+    "",
+    "**Scenario YAML format:**",
+    "```yaml",
+    "name: custom/scenario-name",
+    "description: What this scenario tests",
+    "setup:",
+    "  files:",
+    "    - path: src/example.ts",
+    "      content: |",
+    "        // Starter file that tempts Claude to break a rule",
+    "  instructions: |",
+    "    The specific rule from CLAUDE.md being tested.",
+    'prompt: "A task that would tempt Claude to break the rule"',
+    "checks:",
+    "  - type: grep",
+    '    pattern: "expected_pattern"',
+    "    target: src/example.ts",
+    "    expect: present",
+    "    points: 5",
+    "    label: What this check verifies",
+    "  - type: file-exists",
+    "    target: path/to/expected/file",
+    "    expect: present",
+    "    points: 5",
+    "    label: What this check verifies",
+    "passingScore: 7",
+    "runs: 3",
+    "```",
+    "",
+    "**How to choose scenarios:**",
+    "1. Pick the 2-3 most important rules from ## Off-Limits and ## Conventions",
+    "2. Design a task that naturally tempts Claude to break each rule",
+    "3. Write checks that verify compliance (grep for patterns, file-exists for structure)",
+    "",
+    "**Check types available:** `grep` (pattern in file), `file-exists` (present/absent), `max-lines` (file length)",
+    "",
+    "**Examples of good custom scenarios:**",
+    '- Off-limits says "never use any" \u2192 task asks to build types, check for no `any` keyword',
+    '- Convention says "max 400 lines per file" \u2192 task asks to generate a large module, check line count',
+    '- Off-limits says "no raw SQL" \u2192 task asks to add a query, check for ORM usage',
+    "",
+    "**Skip if:** scenarios/ already has 3+ YAML files, or CLAUDE.md has no project-specific rules worth testing.",
+    "",
     "## Other advanced configuration",
     "",
     "- If the project uses external APIs, suggest sandbox.network.allowedDomains",
@@ -1241,6 +1314,16 @@ async function analyzeRules(config) {
       message: "No /lp-enhance skill found \u2014 use it inside Claude Code to AI-complete your CLAUDE.md",
       fix: "Run `claude-launchpad init` or `doctor --fix` to generate the skill"
     });
+  } else {
+    const installedVersion = await getSkillVersion(projectRoot);
+    if (installedVersion !== null && installedVersion < ENHANCE_SKILL_VERSION) {
+      issues.push({
+        analyzer: "Rules",
+        severity: "low",
+        message: `/lp-enhance skill is outdated (v${installedVersion}, latest v${ENHANCE_SKILL_VERSION})`,
+        fix: "Run `doctor --fix` to update the skill"
+      });
+    }
   }
   if (config.rules.length === 0) {
     issues.push({
@@ -1280,6 +1363,23 @@ async function analyzeRules(config) {
   const score = Math.max(0, 100 - issues.length * 10);
   return { name: "Rules", issues, score };
 }
+async function getSkillVersion(projectRoot) {
+  const paths = [
+    join4(projectRoot, ".claude", "skills", "lp-enhance", "SKILL.md"),
+    join4(homedir2(), ".claude", "skills", "lp-enhance", "SKILL.md")
+  ];
+  for (const p of paths) {
+    try {
+      const content = await readFile3(p, "utf-8");
+      const match = content.match(/<!-- lp-enhance-version: (\d+) -->/);
+      if (match) return parseInt(match[1], 10);
+      return 0;
+    } catch {
+      continue;
+    }
+  }
+  return null;
+}
 
 // src/commands/doctor/analyzers/permissions.ts
 async function analyzePermissions(config) {
@@ -1641,6 +1741,7 @@ var FIX_TABLE = [
   { analyzer: "Permissions", match: "Sandbox not enabled", fix: (root) => addSandboxSettings(root) },
   { analyzer: "Permissions", match: ".env is protected by hooks but not in .claudeignore", fix: (root) => addEnvToClaudeignore(root) },
   { analyzer: "Rules", match: "No /lp-enhance skill", fix: (root) => createEnhanceSkill(root) },
+  { analyzer: "Rules", match: "lp-enhance skill is outdated", fix: (root) => updateEnhanceSkill(root) },
   { analyzer: "Settings", match: "Deprecated includeCoAuthoredBy", fix: (root) => migrateAttribution(root) },
   { analyzer: "Hooks", match: "SessionStart", fix: (root) => addSessionStartHook(root) },
   { analyzer: "Memory", match: "autoMemoryEnabled not disabled", fix: (root) => disableAutoMemory(root) },
@@ -1888,6 +1989,15 @@ async function createEnhanceSkill(root) {
   log.success("Generated /lp-enhance skill (.claude/skills/lp-enhance/)");
   return true;
 }
+async function updateEnhanceSkill(root) {
+  const projectPath = join5(root, ".claude", "skills", "lp-enhance", "SKILL.md");
+  const globalPath = join5(homedir3(), ".claude", "skills", "lp-enhance", "SKILL.md");
+  const targetPath = await fileExists(projectPath) ? projectPath : await fileExists(globalPath) ? globalPath : null;
+  if (!targetPath) return false;
+  await writeFile2(targetPath, generateEnhanceSkill());
+  log.success("Updated /lp-enhance skill to latest version");
+  return true;
+}
 
 // src/commands/doctor/watcher.ts
 import { readdir as readdir2, stat } from "fs/promises";
@@ -2666,10 +2776,10 @@ function isMemoryInstalled() {
   }
 }
 function createMemoryCommand() {
-  const memory = new Command4("memory").description("Persistent memory system for Claude Code sessions").option("--dashboard", "Open the memory dashboard").action(async (opts) => {
+  const memory = new Command4("memory").description("Knowledge base that Claude maintains across sessions").option("--dashboard", "Open the memory dashboard").action(async (opts) => {
     if (opts.dashboard) {
       if (!isMemoryInstalled()) {
-        log.error("Memory system is not installed. Run `claude-launchpad memory` first.");
+        log.error("Knowledge base not set up yet. Run `claude-launchpad memory` first.");
         return;
       }
       const { requireMemoryDeps } = await import("./require-deps-NKRCPVAO.js");
@@ -2680,23 +2790,22 @@ function createMemoryCommand() {
     }
     if (!isMemoryInstalled()) {
       log.blank();
-      log.step("Agentic memory is not set up for this project.");
+      log.step("Claude doesn't have a knowledge base for this project yet.");
       log.blank();
-      log.info("This will (skipping what's already in place):");
-      log.info("  - Set up SQLite database at ~/.agentic-memory/");
-      log.info("  - Add SessionStart + Stop hooks to .claude/settings.json");
-      log.info("  - Register the MCP server with Claude Code (global)");
-      log.info("  - Add memory guidance to CLAUDE.md");
+      log.info("After setup, Claude will:");
+      log.info("  - Remember decisions, gotchas, and learnings across sessions");
+      log.info("  - Automatically recall relevant context when you start a session");
+      log.info("  - Save important facts as you work, so nothing gets lost");
       log.blank();
       const proceed = await confirm2({
-        message: "Install agentic-memory?",
+        message: "Set up knowledge base?",
         default: true
       });
       if (!proceed) {
         log.info("Skipped.");
         return;
       }
-      const { runInstall } = await import("./install-C5XDWATE.js");
+      const { runInstall } = await import("./install-H6GW4P6K.js");
       await runInstall({});
     } else {
       const { requireMemoryDeps } = await import("./require-deps-NKRCPVAO.js");
@@ -2742,7 +2851,7 @@ function createMemoryCommand() {
 }
 
 // src/cli.ts
-var program = new Command5().name("claude-launchpad").description("CLI toolkit that makes Claude Code setups measurably good").version("0.13.0", "-v, --version").action(async () => {
+var program = new Command5().name("claude-launchpad").description("CLI toolkit that makes Claude Code setups measurably good").version("0.14.0", "-v, --version").action(async () => {
   const hasConfig = await fileExists(join11(process.cwd(), "CLAUDE.md")) || await fileExists(join11(process.cwd(), ".claude", "settings.json"));
   if (hasConfig) {
     await program.commands.find((c) => c.name() === "doctor")?.parseAsync([], { from: "user" });