@trac3r/oh-my-god 2.2.11

package/commands/OMG:compat.md

@@ -0,0 +1,57 @@
+---
+description: Run legacy skill names on OMG standalone via compatibility dispatcher.
+allowed-tools: Read, Grep, Glob, Bash(python3:*), Bash(rg:*), Bash(find:*), Bash(cat:*)
+argument-hint: "<skill-name> [optional problem]"
+---
+
+# /OMG:compat
+
+Use this when migrating legacy workflows to OMG standalone.
+
+```bash
+OMG_CLI="${OMG_CLI_PATH:-$HOME/.claude/omg-runtime/scripts/omg.py}"
+if [ ! -f "$OMG_CLI" ] && [ -f "scripts/omg.py" ]; then OMG_CLI="scripts/omg.py"; fi
+```
+
+## List supported skills
+
+```bash
+python3 "$OMG_CLI" compat list
+```
+
+## Inspect contracts
+
+```bash
+python3 "$OMG_CLI" compat contract --all
+python3 "$OMG_CLI" compat contract --skill omg-teams
+python3 "$OMG_CLI" compat snapshot --output runtime/omg_compat_contract_snapshot.json
+python3 scripts/check-omg-compat-contract-snapshot.py --strict-version
+python3 scripts/check-omg-standalone-clean.py
+```
+
+## Generate compatibility gap report
+
+```bash
+python3 "$OMG_CLI" compat gap-report
+```
+
+## Enforce GA gate (CI/local)
+
+```bash
+python3 "$OMG_CLI" compat gate --max-bridge 0
+python3 "$OMG_CLI" compat gate --max-bridge 0 --output .omg/evidence/omg-compat-gap.json
+```
+
+## Run a legacy skill
+
+```bash
+python3 "$OMG_CLI" compat run --skill "<skill-name>" --problem "$ARGUMENTS"
+```
+
+Examples:
+
+```bash
+python3 "$OMG_CLI" compat run --skill omg-teams --problem "review auth flow"
+python3 "$OMG_CLI" compat run --skill plan --problem "ship secure release"
+python3 "$OMG_CLI" compat run --skill pipeline --problem "train model"
+```

package/commands/OMG:cost.md

@@ -0,0 +1,181 @@
+---
+description: "Display session cost tracking, budget status, and usage history."
+allowed-tools: Read, Bash(python*:*), Grep
+argument-hint: "[history|budget|reset]"
+---
+
+# /OMG:cost — Cost Tracker
+
+Display session cost tracking, budget status, and usage history.
+
+## Usage
+
+```
+/OMG:cost
+/OMG:cost history
+/OMG:cost budget
+/OMG:cost reset
+```
+
+## Sub-Commands
+
+### `/OMG:cost` (default)
+
+Show current session cost summary: total tokens, total USD spent, and remaining budget percentage.
+
+```python
+from hooks._cost_ledger import read_cost_summary
+summary = read_cost_summary(".")
+print(f"Total tokens: {summary['total_tokens']}")
+print(f"Total cost: ${summary['total_cost_usd']:.4f}")
+print(f"Entries: {summary['entry_count']}")
+```
+
+### `/OMG:cost history`
+
+Show cost breakdown by tool and by session. Lists top consumers and per-session aggregates.
+
+```python
+from hooks._cost_ledger import read_cost_summary
+summary = read_cost_summary(".")
+
+# By tool
+for tool, stats in sorted(summary["by_tool"].items(), key=lambda x: x[1]["cost_usd"], reverse=True):
+    print(f"  {tool}: {stats['count']} calls, {stats['tokens']} tokens, ${stats['cost_usd']:.4f}")
+
+# By session
+for sid, stats in summary["by_session"].items():
+    print(f"  {sid}: {stats['count']} calls, ${stats['cost_usd']:.4f}")
+```
+
+### `/OMG:cost budget`
+
+Show budget configuration and threshold status. Reads from `_omg.cost_budget` in `settings.json`.
+
+Displays:
+- Session budget limit (USD)
+- Current spend vs limit
+- Remaining budget percentage
+- Threshold levels and which have been triggered
+
+```python
+import json
+with open("settings.json") as f:
+    config = json.load(f)
+budget = config.get("_omg", {}).get("cost_budget", {})
+print(f"Session limit: ${budget.get('session_limit_usd', 5.0):.2f}")
+print(f"Thresholds: {budget.get('thresholds', [50, 80, 95])}")
+print(f"Pricing: ${budget.get('pricing', {}).get('input_per_mtok', 3.0)}/Mtok in, ${budget.get('pricing', {}).get('output_per_mtok', 15.0)}/Mtok out")
+```
+
+### `/OMG:cost reset`
+
+Clear the cost ledger and threshold state for a fresh start.
+
+Removes:
+- `.omg/state/ledger/cost-ledger.jsonl`
+- `.omg/state/.cost-threshold-state.json`
+
+## Feature Flag
+
+- **Flag name**: `OMG_COST_TRACKING_ENABLED`
+- **Default**: `False` (disabled)
+- **Enable**: `export OMG_COST_TRACKING_ENABLED=1`
+
+Or set in `settings.json`:
+
+```json
+{
+  "_omg": {
+    "features": {
+      "COST_TRACKING": true
+    }
+  }
+}
+```
+
+## Budget Configuration
+
+Configure in `settings.json` under `_omg.cost_budget`:
+
+```json
+{
+  "_omg": {
+    "cost_budget": {
+      "session_limit_usd": 5.0,
+      "thresholds": [50, 80, 95],
+      "pricing": {
+        "input_per_mtok": 3.0,
+        "output_per_mtok": 15.0
+      }
+    }
+  }
+}
+```
+
+| Setting | Default | Description |
+|---------|---------|-------------|
+| `session_limit_usd` | `5.0` | Maximum spend per session in USD |
+| `thresholds` | `[50, 80, 95]` | Budget percentage thresholds for alerts |
+| `pricing.input_per_mtok` | `3.0` | Cost per million input tokens |
+| `pricing.output_per_mtok` | `15.0` | Cost per million output tokens |
+
+## Output Example
+
+```
+============================================================
+  OMG Cost Tracker — Session Summary
+============================================================
+
+  Total tokens:     124,350
+  Total cost:       $0.4821
+  Budget remaining: 90.4% ($4.52 of $5.00)
+  Tool calls:       47
+
+  Top consumers:
+    Bash:           23 calls, 68,200 tokens, $0.2644
+    Read:           12 calls, 34,100 tokens, $0.1322
+    Write:           8 calls, 15,050 tokens, $0.0583
+    Edit:            4 calls,  7,000 tokens, $0.0272
+
+============================================================
+```
+
+## Hook Integration
+
+The `budget_governor.py` PostToolUse hook automatically tracks costs when `COST_TRACKING` is enabled. It:
+
+1. Estimates token usage for each tool call (Tier 2 calibrated model)
+2. Appends cost entries to `.omg/state/ledger/cost-ledger.jsonl`
+3. Injects budget status into Claude's context via `additionalContext`
+4. Fires threshold alerts at 50%, 80%, and 95% budget usage
+
+## Safety
+
+- **Read-only** (default): `/cost`, `/cost history`, `/cost budget` only read data
+- **Feature-gated**: Hook and command require `COST_TRACKING` flag enabled
+- **Advisory-only**: Budget governor never blocks tool execution
+- **Crash-isolated**: All operations exit 0 on failure
+
+## API
+
+```python
+from hooks._cost_ledger import read_cost_summary, append_cost_entry, rotate_cost_ledger
+
+# Get aggregated cost summary
+summary = read_cost_summary(".")
+
+# Append a cost entry
+append_cost_entry(".", {
+    "ts": "2026-03-04T12:00:00Z",
+    "tool": "Bash",
+    "tokens_in": 500,
+    "tokens_out": 200,
+    "cost_usd": 0.0045,
+    "model": "claude-opus-4-6",
+    "session_id": "ses_abc123"
+})
+
+# Rotate ledger when > 5MB
+rotate_cost_ledger(".")
+```

package/commands/OMG:crazy.md

@@ -0,0 +1,125 @@
+---
+description: "CRAZY mode — maximum multi-agent orchestration. Claude orchestrates, Codex deep-codes, Gemini designs. Anti-hallucination + error-loop prevention."
+allowed-tools: Read, Write, Edit, MultiEdit, Bash, Grep, Glob, Task
+argument-hint: "[task description]"
+---
+
+# /OMG:crazy — All-Agent Maximum Orchestration
+
+## Phase 1: Intent Classification (BEFORE acting)
+
+STOP. Before doing ANYTHING, say out loud:
+"I understand you want me to [INTENT]: [specific goal]."
+
+| Signal | Intent | Action |
+|--------|--------|--------|
+| fix/bug/error | **FIX** | Debug → patch source → verify |
+| build/create/add | **IMPLEMENT** | Plan → code → test |
+| refactor/clean | **REFACTOR** | Preserve behavior, verify before+after |
+| review/check | **REVIEW** | Read ALL → report → don't change |
+
+If the user hasn't confirmed, ASK. Don't assume.
+
+## Phase 1.5: Brainstorm Merge (AUTOMATIC)
+
+CRAZY mode now includes brainstorming by default. Do not run a separate brainstorm command.
+
+- Generate 2-3 viable approaches quickly
+- Compare trade-offs (risk, effort, reversibility)
+- Pick one approach and proceed immediately
+
+## Phase 2: Agent Dispatch (Parallel Sub-Agent Required)
+
+CRAZY mode must use parallel sub-agent dispatch for worker tracks.
+
+### Mandatory parallel launch pattern
+
+```python
+task(subagent_type="explore", run_in_background=true, load_skills=[], description="Architect track", prompt="...")
+task(subagent_type="explore", run_in_background=true, load_skills=[], description="Backend track", prompt="...")
+task(subagent_type="explore", run_in_background=true, load_skills=[], description="Frontend track", prompt="...")
+task(subagent_type="explore", run_in_background=true, load_skills=[], description="Security track", prompt="...")
+task(subagent_type="explore", run_in_background=true, load_skills=[], description="Verification track", prompt="...")
+```
+
+Then:
+- collect each task with `background_output(task_id="...")`
+- run a `sequential-thinking` synthesis pass to merge contradictions and produce one dependency-ordered checklist
+- only then start implementation
+
+### Claude (You — Orchestrator)
+- Break task into parallel subtasks
+- Delegate to specialists
+- Synthesize results
+- Make architecture decisions
+- Run a sequential-thinking merge before implementation
+
+### Codex — Deep Worker + Code Reviewer
+**Codex=deep-code: backend logic, security, debugging, algorithms, performance, root cause analysis**
+Use for: complex multi-file changes, security audits, root cause analysis, backend architecture
+```
+/OMG:escalate codex "
+Task: [specific implementation task]
+Context: [file paths, patterns to follow]
+Verify: [how to confirm correctness]
+"
+```
+
+### Mandatory Post-Plan Codex Validation
+After any planning step inside CRAZY mode, run a Codex validation pass before implementation:
+
+```bash
+/OMG:escalate codex "Validate this plan for gaps, ordering risks, hidden edge cases, and missing verification. Return only actionable corrections."
+```
+
+### Gemini — UI/UX + Visual Design
+**Gemini=UI/UX: frontend, visual, accessibility, responsive design, CSS, component styling**
+Use for: frontend components, styling, visual review, accessibility, responsive layouts
+```
+/OMG:escalate gemini "
+Task: [UI/design task]
+Context: [component paths, design system]
+"
+```
+
+## Phase 3: Anti-Hallucination Protocol
+
+After EVERY implementation step:
+1. **Run it** — show build/test output with exit code
+2. **Never claim "looks correct"** — evidence or it didn't happen
+3. **Never edit tests to match bugs** — fix the SOURCE code
+4. **Never change only configs/scripts** — verify source files changed
+
+## Phase 4: Error Loop Prevention
+
+```
+IF same_error >= 3:
+    STOP. Do not retry.
+    OPTIONS:
+    1. /OMG:escalate codex "debug: [error] in [file]"
+    2. Completely different approach
+    3. Ask user for guidance
+    NEVER retry the same thing a 4th time.
+```
+
+## Phase 5: Completion Gate
+
+Don't stop until ALL verified:
+- [ ] Changes compile/build (exit 0)
+- [ ] Tests pass (exit 0)
+- [ ] Security check on auth/payment/database code
+- [ ] No TODO/FIXME left in new code
+- [ ] Diff is surgical (minimal, no unnecessary changes)
+
+## Phase 6: Report
+```
+CRAZY Mode — Complete
+━━━━━━━━━━━━━━━━━━━━
+Intent: [what was requested]
+Result: [what was delivered]
+Claude: [orchestration decisions]
+Codex: [deep work done]
+Gemini: [visual work done]
+Verified: [build ✅ | test ✅ | lint ✅]
+Files: [N files, M lines changed]
+```

package/commands/OMG:create-agent.md

@@ -0,0 +1,183 @@
+---
+description: "Wizard command for creating new custom agents in ~/.omg/agents/ or .omg/agents/."
+allowed-tools: Read, Write, Edit, Bash
+argument-hint: "[agent-name]"
+---
+
+# /OMG:create-agent — Custom Agent Creation Wizard
+
+Create a custom agent for your project or user-level configuration.
+
+## Prerequisites
+
+Enable the custom agents feature:
+
+```bash
+export OMG_CUSTOM_AGENTS_ENABLED=1
+```
+
+Or add to your project's `settings.json`:
+
+```json
+{
+  "_omg": {
+    "features": {
+      "CUSTOM_AGENTS": true
+    }
+  }
+}
+```
+
+## Agent Locations
+
+- **User-level**: `~/.omg/agents/<name>.md` — available in all projects
+- **Project-level**: `.omg/agents/<name>.md` — available in this project only
+
+Project-level agents override user-level agents with the same name.
+
+## Quick Start
+
+1. Create the agents directory:
+
+```bash
+# For project-level agents:
+mkdir -p .omg/agents
+
+# For user-level agents:
+mkdir -p ~/.omg/agents
+```
+
+2. Create your agent file (e.g., `.omg/agents/my-agent.md`):
+
+```markdown
+---
+name: my-agent
+description: Brief description of what this agent does
+model: claude-sonnet-4-5
+tools: Read, Grep, Glob, Edit, Write
+bundled: false
+---
+
+# Agent: My Agent
+
+## Role
+
+Describe the agent's primary role and responsibilities here.
+This should be a clear, concise statement of what the agent does.
+
+## Model
+
+`default` (claude-sonnet-4-5) — general-purpose model for this agent.
+
+Available roles: `smol` (haiku, fast), `default` (sonnet), `slow` (opus, deep reasoning).
+
+## Capabilities
+
+- List specific capabilities here
+- What tools does this agent use?
+- What domains does it specialize in?
+
+## Instructions
+
+Detailed behavioral instructions for the agent.
+
+**Core rules:**
+- Rule 1
+- Rule 2
+- Rule 3
+
+**Strategy:**
+1. Step 1
+2. Step 2
+3. Step 3
+
+## Example Prompts
+
+- "Example prompt 1"
+- "Example prompt 2"
+- "Example prompt 3"
+```
+
+## Required Sections
+
+Your agent **must** include these sections to pass validation:
+
+| Section | Required | Description |
+|---------|----------|-------------|
+| `# Agent: <name>` | ✅ Yes | Agent header with name |
+| `## Role` | ✅ Yes | Primary role description |
+| `## Model` | Optional | Model preference (smol/default/slow) |
+| `## Capabilities` | Optional | List of capabilities |
+| `## Instructions` | Optional | Behavioral instructions |
+
+## Validation
+
+Custom agents are validated on load. Invalid agents (missing required sections) are skipped with warnings.
+
+To verify your agent is valid:
+
+```bash
+export OMG_CUSTOM_AGENTS_ENABLED=1
+python3 -c "
+from runtime.custom_agent_loader import load_custom_agents
+agents = load_custom_agents('.')
+for a in agents:
+    status = '✅' if a['validated'] else '❌'
+    print(f\"{status} {a['name']} ({a['level']}) — {a['description'][:60]}\")
+    if a.get('issues'):
+        for issue in a['issues']:
+            print(f\"   ⚠️  {issue}\")
+"
+```
+
+## Examples
+
+### Minimal Valid Agent
+
+```markdown
+# Agent: Greeter
+
+## Role
+
+Simple greeting agent that welcomes users.
+```
+
+### Full Agent with All Sections
+
+See the template in Quick Start above.
+
+### Specialized Domain Agent
+
+```markdown
+# Agent: Data Pipeline
+
+## Role
+
+ETL pipeline specialist. Designs and optimizes data transformation workflows.
+
+## Model
+
+`slow` (claude-opus-4-5) — deep reasoning for complex pipeline design.
+
+## Capabilities
+
+- Design ETL pipelines with error handling and retry logic
+- Optimize SQL queries for large datasets
+- Schema migration planning
+- Data quality validation rules
+
+## Instructions
+
+You are a data engineering specialist.
+
+**Core rules:**
+- Always consider idempotency in pipeline design
+- Prefer incremental processing over full reloads
+- Include monitoring and alerting in every pipeline
+
+**Strategy:**
+1. Understand the data sources and sinks
+2. Design the transformation logic
+3. Add error handling and retry mechanisms
+4. Plan for monitoring and observability
+```

package/commands/OMG:deep-plan.md

@@ -0,0 +1,18 @@
+---
+description: "Strategic planning via the canonical plan-council bundle."
+allowed-tools: Read, Grep, Glob
+argument-hint: "<planning goal or feature to plan>"
+---
+
+# /OMG:deep-plan
+
+`/OMG:deep-plan` is a compatibility path to the canonical `plan-council` bundle.
+
+Users invoke `/OMG:deep-plan`; the runtime routes to `plan-council` for execution.
+For the full behavior specification, see `plugins/advanced/commands/OMG:deep-plan.md`.
+
+## Compatibility Notes
+
+- This alias exists so users can invoke deep-plan workflows from the root command surface.
+- Runtime behavior resolves to the `plan-council` bundle — same structured deliberation, same artifacts.
+- OMG uses the `plan-council` bundle for canonical planning, dissent, and evidence workflows.