azclaude-copilot 0.5.6 → 0.5.7

package/.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
     {
       "name": "azclaude",
       "description": "AZCLAUDE is a complete AI coding environment for Claude Code. It installs 39 commands, 10 auto-invoked skills, 15 specialized agents, 4 hooks, and a persistent memory system — in one command.\n\nKey features:\n• Memory across sessions — goals.md + checkpoints injected automatically before every session\n• Self-improving loop — /reflect fixes stale CLAUDE.md rules, /reflexes learns from tool-use patterns, /evolve creates agents from git evidence\n• Autonomous copilot mode — /copilot runs a three-tier team (orchestrator → problem-architect → milestone-builder) across sessions until the product ships\n• Spec-driven workflow — /constitute writes project rules, /spec writes structured ACs, /analyze detects plan drift and ghost milestones, /blueprint traces every milestone to a spec\n• Security layer — 111-rule environment scan (/sentinel), pre-write secret blocking, pre-ship credential audit\n• Progressive levels 0–10 — start with CLAUDE.md, grow into multi-agent pipelines and self-evolving environments\n• Zero dependencies — no npm packages, no external APIs, no vector databases. Plain markdown files and Claude Code's native architecture.\n• Smart install — npx azclaude-copilot@latest auto-detects first install vs upgrade vs verify. Context-aware onboarding shows the right next command for your project state.\n\nExample use cases:\n• /setup — scan an existing project, detect stack + domain + scale, fill CLAUDE.md, generate project-specific skills and agents automatically\n• /copilot \"Build a compliance SaaS with trilingual support\" — walk away, come back to working code across multiple sessions\n• /sentinel — run a scored security audit (0–100, grade A–F) across hooks, permissions, MCP servers, agent configs, and secrets\n• /evolve — detect gaps in the environment, generate new skills and agents from git co-change evidence, report score delta (e.g. 42/100 → 68/100)\n• /constitute — write your project's constitution (non-negotiables, architectural commitments, definition of done) — gates all future AI actions\n• /analyze — cross-artifact consistency check: ghost milestones, spec vs. code drift, unplanned commits\n• /reflect — find stale, missing, or contradicting rules in CLAUDE.md and propose exact fixes\n• /debate \"REST vs GraphQL for this project\" — adversarial evidence-based decision with order-independent scoring, logged to decisions.md",
-      "version": "0.5.6",
+      "version": "0.5.7",
       "source": {
         "source": "github",
         "repo": "haytamAroui/AZ-CLAUDE-COPILOT",

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "azclaude",
-  "version": "0.5.6",
+  "version": "0.5.7",
   "description": "AZCLAUDE is a complete AI coding environment for Claude Code. It installs 39 commands, 10 auto-invoked skills, 15 specialized agents, 4 hooks, and a persistent memory system — in one command.\n\nKey features:\n• Memory across sessions — goals.md + checkpoints injected automatically before every session\n• Self-improving loop — /reflect fixes stale CLAUDE.md rules, /reflexes learns from tool-use patterns, /evolve creates agents from git evidence\n• Autonomous copilot mode — /copilot runs a three-tier team (orchestrator → problem-architect → milestone-builder) across sessions until the product ships\n• Spec-driven workflow — /constitute writes project rules, /spec writes structured ACs, /analyze detects plan drift and ghost milestones, /blueprint traces every milestone to a spec\n• Security layer — 111-rule environment scan (/sentinel), pre-write secret blocking, pre-ship credential audit\n• Progressive levels 0–10 — start with CLAUDE.md, grow into multi-agent pipelines and self-evolving environments\n• Zero dependencies — no npm packages, no external APIs, no vector databases. Plain markdown files and Claude Code's native architecture.\n• Smart install — npx azclaude-copilot@latest auto-detects first install vs upgrade vs verify. Context-aware onboarding shows the right next command for your project state.\n\nExample use cases:\n• /setup — scan an existing project, detect stack + domain + scale, fill CLAUDE.md, generate project-specific skills and agents automatically\n• /copilot \"Build a compliance SaaS with trilingual support\" — walk away, come back to working code across multiple sessions\n• /sentinel — run a scored security audit (0–100, grade A–F) across hooks, permissions, MCP servers, agent configs, and secrets\n• /evolve — detect gaps in the environment, generate new skills and agents from git co-change evidence, report score delta (e.g. 42/100 → 68/100)\n• /constitute — write your project's constitution (non-negotiables, architectural commitments, definition of done) — gates all future AI actions\n• /analyze — cross-artifact consistency check: ghost milestones, spec vs. code drift, unplanned commits\n• /reflect — find stale, missing, or contradicting rules in CLAUDE.md and propose exact fixes\n• /debate \"REST vs GraphQL for this project\" — adversarial evidence-based decision with order-independent scoring, logged to decisions.md",
   "author": {
     "name": "haytamAroui",

package/README.md

@@ -12,6 +12,7 @@
     <a href="#how-it-works-the-execution-pipeline">Pipeline</a> ·
     <a href="#install">Install</a> ·
     <a href="#what-you-get">What You Get</a> ·
+    <a href="#structured-for-claude-frontmatter--xml-tags">Structure</a> ·
     <a href="#architecture-philosophy">Architecture</a> ·
     <a href="#autonomous-mode">Autonomous Mode</a> ·
     <a href="#parallel-execution">Parallel</a> ·
@@ -155,6 +156,87 @@ npx azclaude-copilot@latest doctor   # 32 checks — verify everything is wired
 
 ---
 
+## Structured for Claude: Frontmatter + XML Tags
+
+Anthropic's research shows Claude comprehends structured content significantly better when it uses two patterns natively built into Claude Code: **YAML frontmatter** for metadata and **`<instructions>` XML tags** for content. AZCLAUDE applies both to every agent and skill file — not as convention, but because it measurably changes how Claude reads, routes, and executes them.
+
+### Why this matters
+
+Claude was trained on XML-structured content. Tags like `<instructions>`, `<context>`, and `<examples>` signal boundaries Claude parses reliably — unlike prose headings which are ambiguous. Frontmatter gives Claude Code the metadata it needs to route, describe, and invoke agents without any extra configuration.
+
+### What AZCLAUDE agent files look like
+
+```yaml
+---                                    ← YAML frontmatter block
+name: code-reviewer                    ← agent identifier (used by subagent_type=)
+description: >                         ← Claude Code reads this to decide when to spawn
+  Autonomous code review agent.
+  Use when: review, audit, PR review,
+  find bugs, security check.
+model: opus                            ← which model this agent runs on
+tools: [Read, Glob, Grep, Bash]        ← allowed tools (principle of least privilege)
+disallowedTools: [Write, Edit, Agent]  ← explicit deny list — no write access
+permissionMode: plan                   ← read-only by default
+maxTurns: 30                           ← bounded execution
+tags: [review, pr, quality]            ← semantic routing hints
+---
+
+# Code Reviewer
+
+<instructions>                         ← XML tag: Claude parses this as structured directives
+
+## Layer 1: PERSONA
+Code review specialist. Read-only — never modifies code.
+
+## Layer 2: SCOPE
+- Reviews staged/unstaged changes via `git diff`
+- Checks for bugs, logic errors, edge cases
+- Identifies security issues (injection, XSS, OWASP top 10)
+
+</instructions>
+```
+
+### What AZCLAUDE skill files look like
+
+```yaml
+---                                    ← YAML frontmatter
+name: test-first
+description: >                         ← auto-invocation trigger — Claude reads this
+  Guides test-driven development.
+  Use when about to write code, implement a feature, fix a bug,
+  refactor, add an endpoint, or modify business logic.
+tags: [tdd, testing, coverage]
+---
+
+# Test-First
+
+<instructions>                         ← structured content block
+
+## Check before enforcing
+TDD is opt-in. Check BOTH signals:
+1. CLAUDE.md has a TDD rule
+2. Test files exist (*.test.*, *.spec.*)
+
+Both present → TDD protocol active.
+Either missing → suggest TDD, don't block.
+
+</instructions>
+```
+
+### Why this architecture works
+
+| Pattern | What it does |
+|---------|-------------|
+| `description:` frontmatter | Claude Code reads this to decide when to auto-invoke a skill or which agent to spawn. No code required. |
+| `model: opus` frontmatter | Routes complex review tasks to the most capable model automatically. |
+| `tools:` + `disallowedTools:` | Principle of least privilege — code-reviewer cannot Write, milestone-builder cannot call Agent. |
+| `<instructions>` XML tag | Signals to Claude "parse this as directives" — not prose, not documentation, not commentary. |
+| `permissionMode: plan` | Agents that only read run in plan mode — no accidental writes, no permission prompts. |
+
+All 15 AZCLAUDE agents and all 10 skills use this exact format. Claude Code routes them, describes them in `/help`, and invokes them via `subagent_type=` — all from the frontmatter, zero extra config.
+
+---
+
 ## Architecture Philosophy
 
 **AZCLAUDE uses Markdown files and lifecycle hooks — not MCP servers — as its core architecture.** This is a deliberate engineering decision, not a gap.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "azclaude-copilot",
-  "version": "0.5.6",
+  "version": "0.5.7",
   "description": "AI coding environment — 39 commands, 10 skills, 15 agents, memory, reflexes, evolution. Install: npx azclaude-copilot@latest, then open Claude Code.",
   "bin": {
     "azclaude": "bin/cli.js",

package/templates/hooks/user-prompt.js

@@ -245,9 +245,10 @@ try {
       }
     } else if (pct >= 70) {
       console.log('');
-      console.log(`--- COMPACTION WARNING (${pct}%) ---`);
-      console.log(`Context at ${pct}% — compaction approaching. Run /snapshot to save session state.`);
-      console.log('--- END WARNING ---');
+      console.log(`!!! SNAPSHOT REQUIRED (${pct}% context used) !!!`);
+      console.log(`Run /snapshot NOW — Claude will compact and lose your session reasoning soon.`);
+      console.log(`This warning repeats every message until you run /snapshot or context resets.`);
+      console.log(`!!! /snapshot !!! /snapshot !!! /snapshot !!!`);
     }
   }
 } catch (_) {}