orbital-command 0.1.0

package/templates/agents/README.md

@@ -0,0 +1,188 @@
+---
+name: agent-system
+description: Overview of the Orbital Command agent system. Load when understanding how agents work together.
+tokens: ~2K
+load-when: Understanding the agent system
+last-verified: 2026-01-11
+---
+
+# Orbital Command Agent System
+
+## Overview
+
+This project uses an automated multi-agent system to ensure code quality, security, and consistency. Agents are specialized AI reviewers that automatically activate based on what you're working on.
+
+**You don't invoke agents manually.** The system automatically determines which agents are needed based on:
+1. Files you're touching (hard triggers)
+2. Type of task you're doing (smart triggers)
+3. Risk level of changes (security mode)
+
+---
+
+## ⛔ Anti-Bypass Protocol
+
+**All agents MUST follow these rules when blocked by enforcement:**
+
+### When Blocked, You MUST:
+
+1. **READ the error message** - It contains the solution
+2. **FIX the underlying issue** - Don't look for shortcuts
+3. **ASK the user** - If genuinely stuck after trying the suggested fix
+
+### When Blocked, You MUST NOT:
+
+| ❌ Forbidden Action | Why It's Wrong |
+|---------------------|----------------|
+| `cat > file.ts << 'EOF'` | Bypasses Write tool - use Write properly |
+| `cp /tmp/file.ts src/` | Workaround for blocked Write |
+| `sed -i 's/x/y/' file.ts` | Bypasses Edit tool - use Edit properly |
+| `git commit --no-verify` | Skips quality checks that exist for a reason |
+| Giving up after one block | The error message tells you how to proceed |
+
+### Multi-Layer Enforcement
+
+Bypass attempts are caught at multiple levels:
+
+```
+PreToolUse Hook  →  Blocks bash workarounds BEFORE execution
+       ↓
+Pre-commit Hook  →  Catches workarounds in staged changes
+       ↓
+CI Pipeline      →  Final gate catches anything that slips through
+       ↓
+Audit Database   →  All attempts logged for pattern analysis
+```
+
+### Why This Matters
+
+Every enforcement rule exists because a past mistake taught us it matters. Bypassing enforcement just delays problems until they cause production incidents.
+
+**If a rule is genuinely wrong:** Discuss with the user - don't work around it.
+
+---
+
+## The Team
+
+```
+┌─────────────────────────────────────────────────────────────────────────────┐
+│                                                                             │
+│                      🏛️  ORBITAL COMMAND AGENT TEAM                        │
+│                                                                             │
+├─────────────────────────────────────────────────────────────────────────────┤
+│                                                                             │
+│  🔴 RED TEAM (Adversarial)                                                  │
+│  "Find problems before users do"                                           │
+│                                                                             │
+│     🗡️ Attacker        - Exploits security vulnerabilities                  │
+│     💥 Chaos Agent     - Imagines failure modes and edge cases             │
+│                                                                             │
+│  🔵 BLUE TEAM (Domain Experts)                                              │
+│  "Deep knowledge, consulted on demand"                                     │
+│                                                                             │
+│     🎨 Frontend Designer - Components, UX, style consistency               │
+│     (Add domain experts for your project)                                 │
+│                                                                             │
+│  🟢 GREEN TEAM (Guardians)                                                  │
+│  "Protect quality and standards"                                           │
+│                                                                             │
+│     🏗️ Architect       - Patterns, structure, module boundaries            │
+│     📋 Rules Enforcer  - Non-negotiable project rules (automated)          │
+│                                                                             │
+└─────────────────────────────────────────────────────────────────────────────┘
+```
+
+---
+
+## How It Works
+
+### 1. Task Triage (Automatic)
+
+When you describe a task, the system automatically:
+- Identifies which files will likely be affected
+- Determines which agents should review
+- Spawns relevant agents in parallel
+
+### 2. Agent Review
+
+Each triggered agent:
+- Loads its specific context (domain docs, patterns)
+- Reviews the task/implementation from its perspective
+- Reports findings with severity levels
+
+### 3. Synthesis
+
+All agent outputs are combined into:
+- **Consensus**: What all agents agree on
+- **Concerns**: Issues raised (with severity)
+- **Conflicts**: Where agents disagree (escalated to you if major)
+
+### 4. Quality Gate
+
+Before any commit:
+- 📋 Rules Enforcer runs automatically
+- Must pass all checks to proceed
+
+---
+
+## Severity Levels
+
+| Level | Icon | Meaning | Action |
+|-------|------|---------|--------|
+| BLOCKER | 🚫 | Cannot proceed | Must fix before continuing |
+| WARNING | ⚠️ | Should fix | Can proceed with caution |
+| SUGGESTION | 💡 | Consider | Nice to have, not required |
+| PASSED | ✅ | Verified OK | Explicitly checked and good |
+
+---
+
+## Invocation Modes
+
+### Full Mode (Default)
+- All triggered agents run
+- Complete synthesis
+- Use for: Features, refactors, significant changes
+
+### Quick Mode
+- Only 📋 Rules Enforcer runs
+- Request with: "quick mode: [task]"
+- Use for: Typos, comments, tiny fixes
+- **Cannot use for**: Security files, new endpoints
+
+### Security Mode (Automatic)
+- 🗡️ Attacker runs twice (before AND after implementation)
+- Auto-triggered for: Auth, encryption, sensitive operations, user input
+
+---
+
+## Conflict Resolution
+
+**Minor conflicts** → Auto-resolved using priority hierarchy:
+1. 🗡️ Attacker (security wins)
+2. 💥 Chaos Agent (reliability)
+3. Domain experts (correctness)
+4. 🏗️ Architect (patterns)
+5. 🎨 Frontend Designer (aesthetics)
+
+**Major conflicts** → Escalated to you with recommendation
+
+---
+
+## Quick Reference
+
+| I'm doing... | Agents auto-triggered |
+|--------------|----------------------|
+| New feature | 🏗️ + 💥 + 🎨 + [domain experts] |
+| Bug fix | 💥 + [domain expert] |
+| New API endpoint | 🗡️ + 🏗️ |
+| Frontend work | 🎨 + 🏗️ |
+| Deployment/CI changes | 🚀 + 💥 |
+| Database migrations | 🚀 + 🏗️ |
+| Security-sensitive | 🗡️ + 💥 (ALWAYS) |
+| Any commit | 📋 (ALWAYS) |
+
+---
+
+## Related Files
+
+- [AUTO-INVOKE.md](./AUTO-INVOKE.md) - Detailed trigger rules
+- [CONFLICT-RESOLUTION.md](./CONFLICT-RESOLUTION.md) - How conflicts are handled

package/templates/agents/SKILL-TRIGGERS.md

@@ -0,0 +1,100 @@
+---
+name: skill-triggers
+description: Defines when to suggest or auto-invoke skills based on user phrases and context.
+tokens: ~1.5K
+load-when: Loaded alongside AUTO-INVOKE.md during task triage
+last-verified: 2026-03-17
+---
+
+# Skill Auto-Triggers
+
+Skills should be suggested proactively based on user language patterns and session context.
+
+---
+
+## Phrase-Based Triggers
+
+### Scope Lifecycle
+
+| User Says | Suggest Skill | Why |
+|-----------|---------------|-----|
+| "Create a scope for X" | `/scope-create` | New scope document |
+| "Let's plan X" | `/scope-create` | Planning phase |
+| "Implement scope NNN" | `/scope-implement` | Execute scope end-to-end |
+| "Work on scope NNN" | `/scope-implement` | Execute scope |
+| "Review scope NNN" | `/scope-pre-review` | Agent team review |
+| "Is scope NNN ready?" | `/scope-verify` | Formal review gate |
+| "Run post-review on NNN" | `/scope-post-review` | Post-implementation review |
+
+### Git & Deployment
+
+| User Says | Suggest Skill | Why |
+|-----------|---------------|-----|
+| "Commit this" | `/git-commit` | Commit to feature branch |
+| "Save our progress" | `/git-commit` | Mid-task checkpoint |
+| "Push to main" | `/git-main` | Push/PR to main |
+| "Merge to dev" | `/git-dev` | Merge into dev branch |
+| "PR to staging" | `/git-staging` | Create staging PR |
+| "Deploy to production" | `/git-production` | Create release PR |
+| "Emergency fix" | `/git-hotfix` | Hotfix workflow |
+
+### Testing & Quality
+
+| User Says | Suggest Skill | Why |
+|-----------|---------------|-----|
+| "Run checks" | `/test-checks` | Quality gates |
+| "Run quality gates" | `/test-checks` | Quality gates |
+| "Code review" | `/test-code-review` | Full code review |
+| "Review the code" | `/test-code-review` | Full code review |
+
+### Session Management
+
+| User Says | Suggest Skill | Why |
+|-----------|---------------|-----|
+| "Where were we?" | `/session-resume` | Resume from interruption |
+| "What was I working on?" | `/session-resume` | Resume from interruption |
+| "Continue where we left off" | `/session-resume` | Resume from interruption |
+| "I'm done for now" | `/git-commit` | Session wrap-up |
+| "That's all for today" | `/git-commit` | Session wrap-up |
+| "Wrap up" | `/git-commit` | Session wrap-up |
+
+---
+
+## Context-Based Auto-Triggers
+
+These are **automatic** suggestions based on session state, not user phrases.
+
+| Context | Auto-Suggest | Rationale |
+|---------|--------------|-----------|
+| Session starts with uncommitted changes | `/session-resume` | User likely continuing previous work |
+| After context compaction | `/session-resume` | Rebuild context efficiently |
+| Before risky/destructive operation | `/git-commit` | Create recovery point |
+| User says "bye/done/that's all" | `/git-commit` | Clean session closure |
+
+---
+
+## Smart Triggers by Task Pattern
+
+| Task Pattern | Skills to Consider |
+|--------------|-------------------|
+| Scope implementation | `/scope-implement` → `/test-checks` → `/git-commit` |
+| Post-implementation review | `/scope-post-review` (chains: `/test-checks` → `/scope-verify` → `/test-code-review`) |
+| Release to staging | `/git-staging` |
+| Release to production | `/git-production` |
+| Emergency fix | `/git-hotfix` |
+
+---
+
+## Priority Order
+
+When multiple skills could apply, prioritize:
+
+1. **Checkpoint before risk**: `/git-commit` before destructive or complex operations
+2. **Quality before merge**: `/test-checks` before `/git-main` or `/git-staging`
+3. **Handoff at session end**: `/git-commit` when user indicates they're done
+
+---
+
+## Related
+
+- [AUTO-INVOKE.md](./AUTO-INVOKE.md) - Agent auto-invocation rules