Optimize CLI version checker workflow based on performance analysis

.github/workflows/cli-version-checker.md

@@ -6,20 +6,25 @@ on:
 permissions:
   contents: read
   actions: read
-engine: claude
+engine:
+  id: claude
+  max-turns: 30
 network: 
    allowed: [defaults, "registry.npmjs.org", "api.github.com", "ghcr.io"]
 imports:
   - shared/jqschema.md
 tools:
   web-fetch:
+  cache-memory: true
   bash:
     - "cat *"
     - "ls *"
     - "grep *"
     - "git *"
     - "make *"
     - "npm install *"
+    - "npm view *"
+    - "npm list *"
     - "claude-code --help"
     - "copilot --help"
     - "codex --help"
@@ -41,17 +46,24 @@ Monitor and update agentic CLI tools: Claude Code, GitHub Copilot CLI, OpenAI Co
 
 ## Process
 
+**EFFICIENCY FIRST**: Before starting:
+1. Check cache-memory at `/tmp/gh-aw/cache-memory/` for previous version checks and help outputs
+2. If cached versions exist and are recent (< 24h), verify if updates are needed before proceeding
+3. If no version changes detected, exit early with success
+
 For each CLI/MCP server:
-1. Fetch latest version from NPM registry or GitHub releases
+1. Fetch latest version from NPM registry or GitHub releases (use npm view commands for package metadata)
 2. Compare with current version in `./pkg/constants/constants.go`
 3. If newer version exists, research changes and prepare update
 
 ### Version Sources
-- **Claude Code**: `https://registry.npmjs.org/@anthropic-ai/claude-code/latest`
-- **Copilot CLI**: `https://registry.npmjs.org/@github/copilot/latest`
-- **Codex**: `https://registry.npmjs.org/@openai/codex/latest`
+- **Claude Code**: Use `npm view @anthropic-ai/claude-code version` (faster than web-fetch)
+- **Copilot CLI**: Use `npm view @github/copilot version`
+- **Codex**: Use `npm view @openai/codex version`
 - **GitHub MCP Server**: `https://api.github.com/repos/github/github-mcp-server/releases/latest`
 
+**Optimization**: Fetch all versions in parallel using multiple npm view or WebFetch calls in a single turn.
+
 ### Research & Analysis
 For each update, analyze intermediate versions:
 - Categorize changes: Breaking, Features, Fixes, Security, Performance
@@ -60,12 +72,17 @@ For each update, analyze intermediate versions:
 - Assign risk level (Low/Medium/High)
 
 ### Tool Installation & Discovery
+**CACHE OPTIMIZATION**: 
+- Before installing, check cache-memory for previous help outputs
+- Only install and run --help if version has changed
+- Store help outputs in cache-memory at `/tmp/gh-aw/cache-memory/[tool]-[version]-help.txt`
+
 For each CLI tool update:
-1. Install the new version globally:
+1. Install the new version globally (skip if already installed from cache check):
    - Claude Code: `npm install -g @anthropic-ai/claude-code@<version>`
    - Copilot CLI: `npm install -g @github/copilot@<version>`
    - Codex: `npm install -g @openai/codex@<version>`
-2. Invoke help to discover commands and flags:
+2. Invoke help to discover commands and flags (compare with cached output if available):
    - Run `claude-code --help`
    - Run `copilot --help`
    - Run `codex --help`
@@ -74,6 +91,7 @@ For each CLI tool update:
    - New command-line flags or options
    - Deprecated or removed features
    - Changed default behaviors
+4. Save new help output to cache-memory for future runs
 
 ### Update Process
 1. Edit `./pkg/constants/constants.go` with new version(s)
@@ -108,12 +126,19 @@ Template structure:
 - Only update stable versions (no pre-releases)
 - Prioritize security updates
 - Document all intermediate versions
+- **USE NPM COMMANDS**: Use `npm view` instead of web-fetch for package metadata queries
+- **CHECK CACHE FIRST**: Before re-analyzing versions, check cache-memory for recent results
+- **PARALLEL FETCHING**: Fetch all versions in parallel using multiple npm/WebFetch calls in one turn
+- **EARLY EXIT**: If no version changes detected, save check timestamp to cache and exit successfully
 - Install and test CLI tools to discover new features via `--help`
 - Compare help output between old and new versions
+- **SAVE TO CACHE**: Store help outputs and version check results in cache-memory
 - Test with `make recompile` before creating PR
 - **DO NOT COMMIT** `*.lock.yml` or `pkg/workflow/js/*.js` files directly
 
 ## Error Handling
+- **SAVE PROGRESS**: Before exiting on errors, save current state to cache-memory
+- **RESUME ON RESTART**: Check cache-memory on startup to resume from where you left off
 - Retry NPM registry failures once after 30s
 - Continue if individual changelog fetch fails
 - Skip PR creation if recompile fails