oh-my-customcode 0.1.2 → 0.1.4

package/templates/agents/manager/creator/refs/rules/MUST-sync-verification.md

@@ -0,0 +1,219 @@
+# [MUST] Sync Verification Rules
+
+> **Priority**: MUST - ENFORCED after any structural changes
+> **ID**: R017
+> **Violation**: Committing without verification = Rule violation
+
+## CRITICAL
+
+**After modifying agents, skills, guides, commands, or pipelines, you MUST run the full verification process before committing.**
+
+```
+╔══════════════════════════════════════════════════════════════════╗
+║  ⚠️  ABSOLUTE PROHIBITION                                        ║
+║                                                                   ║
+║  DO NOT even ASK "커밋하시겠습니까?" after structural changes    ║
+║  until sauron:watch has been executed and passed.                ║
+║                                                                   ║
+║  WRONG:                                                          ║
+║    [Make changes] → "커밋하시겠습니까?"                          ║
+║                                                                   ║
+║  CORRECT:                                                        ║
+║    [Make changes] → sauron:watch → [All pass] → "커밋할까요?"   ║
+║                                                                   ║
+║  Asking to commit before verification = Rule violation           ║
+╚══════════════════════════════════════════════════════════════════╝
+```
+
+```
+╔══════════════════════════════════════════════════════════════════╗
+║  MANDATORY VERIFICATION PROCESS:                                 ║
+║                                                                   ║
+║  Phase 1: Manager Agent Verification (5 rounds)                  ║
+║    - supplier:audit                                              ║
+║    - sync-checker:check                                          ║
+║    - updater:docs                                                ║
+║                                                                   ║
+║  Phase 2: Deep Review (3 rounds)                                 ║
+║    - Workflow alignment                                          ║
+║    - Reference verification                                      ║
+║    - Philosophy compliance                                       ║
+║                                                                   ║
+║  Phase 3: Fix all issues                                         ║
+║                                                                   ║
+║  Phase 4: Commit via gitnerd                                     ║
+║                                                                   ║
+║  Skipping ANY phase = Rule violation                             ║
+╚══════════════════════════════════════════════════════════════════╝
+```
+
+## Phase 1: Manager Agent Verification (5 Rounds)
+
+Run manager agents 5 times to catch all issues:
+
+### Round 1-2: Basic Sync
+```
+□ supplier:audit - Check all agent dependencies
+□ sync-checker:check - Verify registry synchronization
+□ Fix any issues found
+```
+
+### Round 3-4: Deep Sync
+```
+□ supplier:audit - Re-verify after fixes
+□ sync-checker:check - Re-verify registries
+□ updater:docs - Check documentation sync
+□ Fix any remaining issues
+```
+
+### Round 5: Final Verification
+```
+□ All agent counts match (CLAUDE.md, index.yaml, actual files)
+□ All symlinks valid
+□ All command registries updated
+□ All intent-detection triggers present
+```
+
+## Phase 2: Deep Review (3 Rounds)
+
+### Deep Round 1: Workflow Alignment
+```
+□ Agent creation workflow documented and functional
+□ Development workflow uses proper orchestrators
+□ Deployment workflow defined (if applicable)
+□ All orchestrators have complete `manages:` sections
+```
+
+### Deep Round 2: Reference Verification
+```
+□ All orchestrators properly reference their managed agents
+□ All rules are properly referenced
+□ No orphaned agents (not managed by any orchestrator)
+□ No circular references
+```
+
+### Deep Round 3: Philosophy Compliance
+```
+□ R006: Separation of concerns (AGENT.md = role only, no details)
+□ R009: Parallel execution enabled for orchestrators
+□ R010: Multi-agent tasks use orchestrators
+□ R007/R008: Agent/tool identification documented
+□ All MUST rules enforced, SHOULD rules recommended
+```
+
+## Verification Checklist Summary
+
+```
+Phase 1: Manager Verification (5x)
+  □ Round 1: supplier:audit
+  □ Round 2: sync-checker:check
+  □ Round 3: supplier:audit (re-verify)
+  □ Round 4: sync-checker:check + updater:docs
+  □ Round 5: Final count verification
+
+Phase 2: Deep Review (3x)
+  □ Round 1: Workflow alignment
+  □ Round 2: Reference verification
+  □ Round 3: Philosophy compliance
+
+Phase 3: Fix Issues
+  □ All issues from Phase 1 fixed
+  □ All issues from Phase 2 fixed
+  □ Re-run verification if major fixes made
+
+Phase 4: Commit
+  □ Delegate to gitnerd
+  □ Meaningful commit message
+```
+
+## When to Run Full Verification
+
+| Action | Full Verification Required |
+|--------|---------------------------|
+| Create new agent | **YES** |
+| Rename agent | **YES** |
+| Move agent | **YES** |
+| Delete agent | **YES** |
+| Modify agent structure | **YES** |
+| Add/modify skill | **YES** |
+| Add/modify guide | **YES** |
+| Add/modify command | **YES** |
+| Add/modify pipeline | **YES** |
+| Change orchestrator manages section | **YES** |
+| Update rules | **YES** |
+
+## Workflow Diagram
+
+```
+Content Change Made
+        │
+        ▼
+┌───────────────────────────────┐
+│  Phase 1: Manager Verification │
+│  (5 rounds)                    │
+│  supplier → sync-checker →     │
+│  updater                       │
+└───────────────┬───────────────┘
+                │
+                ▼
+┌───────────────────────────────┐
+│  Phase 2: Deep Review          │
+│  (3 rounds)                    │
+│  Workflow → References →       │
+│  Philosophy                    │
+└───────────────┬───────────────┘
+                │
+        ┌───────┴───────┐
+        ▼               ▼
+  [Issues Found]   [All Clean]
+        │               │
+        ▼               │
+   Fix Issues           │
+        │               │
+        └───────┬───────┘
+                │
+                ▼
+┌───────────────────────────────┐
+│  Phase 4: Commit via gitnerd   │
+└───────────────────────────────┘
+```
+
+## Self-Check Before Commit
+
+```
+╔══════════════════════════════════════════════════════════════════╗
+║  BEFORE COMMITTING, ASK YOURSELF:                                ║
+║                                                                   ║
+║  1. Did I complete all 5 rounds of manager verification?         ║
+║  2. Did I complete all 3 rounds of deep review?                  ║
+║  3. Did I fix ALL discovered issues?                             ║
+║  4. Are all counts matching across all sources?                  ║
+║  5. Am I delegating to gitnerd for the commit?                   ║
+║                                                                   ║
+║  If NO to ANY → DO NOT COMMIT                                    ║
+╚══════════════════════════════════════════════════════════════════╝
+```
+
+## Quick Verification Commands
+
+```bash
+# Agent count check
+find agents -name "AGENT.md" | wc -l
+
+# Broken symlinks
+find agents -type l ! -exec test -e {} \; -print
+
+# Index registry count
+grep -c "name:" agents/index.yaml
+
+# Intent trigger coverage
+grep -c "keywords:" skills/orchestration/intent-detection/patterns/agent-triggers.yaml
+```
+
+## Integration with Other Rules
+
+| Rule | Integration |
+|------|-------------|
+| R009 (Parallel) | Verification checks can run in parallel |
+| R010 (Orchestrator) | Delegate commit to gitnerd |
+| R016 (Improvement) | Update this rule if new issues found |

package/templates/agents/manager/creator/refs/rules/MUST-tool-identification.md

@@ -0,0 +1,174 @@
+# [MUST] Tool Usage Identification Rules
+
+> **Priority**: MUST - ENFORCED, NO EXCEPTIONS
+> **ID**: R008
+> **Violation**: Immediate correction required
+
+## CRITICAL
+
+**EVERY tool call MUST be prefixed with agent and model identification. This is NON-NEGOTIABLE.**
+
+```
+Before EVERY tool call, output:
+[agent-name][model] → Tool: <tool-name>
+[agent-name][model] → Target: <file/path/url>
+```
+
+Failure to identify tool usage = Rule violation = Must be corrected immediately.
+
+### Self-Check Before EVERY Tool Call
+
+```
+╔══════════════════════════════════════════════════════════════════╗
+║  BEFORE INVOKING ANY TOOL, ASK YOURSELF:                         ║
+║                                                                   ║
+║  1. Did I output the tool identification line with MODEL?        ║
+║     [agent-name][model] → Tool: <tool-name>                      ║
+║     [agent-name][model] → Target: <target>                       ║
+║                                                                   ║
+║  2. Am I about to call multiple tools in parallel?               ║
+║     → EACH tool needs its own identification line                ║
+║     → List all identifications BEFORE the tool calls             ║
+║                                                                   ║
+║  If NO to #1 → STOP. Output identification FIRST.                ║
+╚══════════════════════════════════════════════════════════════════╝
+```
+
+### Model Values
+
+| Model | When to Use |
+|-------|-------------|
+| `opus` | Complex reasoning, architecture design |
+| `sonnet` | General tasks, code generation (default) |
+| `haiku` | Fast simple tasks, file search |
+
+### Common Violations to Avoid
+
+```
+❌ WRONG: Calling tools without identification
+   "먼저 JD 내용을 확인하겠습니다."
+   <tool_call>WebFetch(...)</tool_call>
+
+❌ WRONG: Missing model in identification
+   [secretary] → Tool: WebFetch
+   [secretary] → Fetching: https://example.com/jd.md
+
+✓ CORRECT: Always identify with agent AND model
+   "먼저 JD 내용을 확인하겠습니다."
+   [secretary][opus] → Tool: WebFetch
+   [secretary][opus] → Fetching: https://example.com/jd.md
+   <tool_call>WebFetch(...)</tool_call>
+
+❌ WRONG: Parallel calls without listing all identifications
+   <tool_call>WebFetch(url1)</tool_call>
+   <tool_call>WebFetch(url2)</tool_call>
+   <tool_call>Bash(cmd)</tool_call>
+
+✓ CORRECT: List all identifications with models, then call
+   [secretary][opus] → Tool: WebFetch
+   [secretary][opus] → Fetching: url1
+   [secretary][opus] → Tool: WebFetch
+   [secretary][opus] → Fetching: url2
+   [secretary][opus] → Tool: Bash
+   [secretary][opus] → Running: cmd
+   <tool_call>WebFetch(url1)</tool_call>
+   <tool_call>WebFetch(url2)</tool_call>
+   <tool_call>Bash(cmd)</tool_call>
+```
+
+## Purpose
+
+Display which agent is using which tool for transparency and debugging. This extends R007 (Agent Identification) to cover tool operations.
+
+## Tool Usage Format
+
+When invoking a tool, prefix with agent and model identification:
+
+```
+[agent-name][model] → Tool: <tool-name>
+```
+
+### Examples
+
+```
+[golang-expert][sonnet] → Tool: Read
+[golang-expert][sonnet] → Reading: src/main.go
+
+[supplier][haiku] → Tool: Glob
+[supplier][haiku] → Searching: agents/**/index.yaml
+
+[creator][sonnet] → Tool: Write
+[creator][sonnet] → Writing: agents/sw-engineer/new-agent/AGENT.md
+```
+
+## Extended Format (Verbose)
+
+For detailed tracking:
+
+```
+┌─ Agent: golang-expert (sw-engineer)
+├─ Model: sonnet
+├─ Skill: go-best-practices
+├─ Tool: Read
+└─ Target: src/main.go
+```
+
+## Tool Categories
+
+| Category | Tools | Format |
+|----------|-------|--------|
+| File Read | Read, Glob, Grep | `→ Reading:` / `→ Searching:` |
+| File Write | Write, Edit | `→ Writing:` / `→ Editing:` |
+| Network | WebFetch | `→ Fetching:` |
+| Execution | Bash, Task | `→ Running:` / `→ Spawning:` |
+
+## When to Display
+
+| Situation | Display |
+|-----------|---------|
+| Reading files | Agent + file path |
+| Writing files | Agent + file path |
+| Searching | Agent + pattern |
+| Fetching URLs | Agent + URL |
+| Running commands | Agent + command |
+
+## Simplified Format
+
+For inline operations:
+
+```
+[supplier][haiku] → Glob: agents/**/index.yaml
+[creator][sonnet] → Write: agents/sw-engineer/new-agent/AGENT.md
+[updater][sonnet] → WebFetch: https://github.com/...
+```
+
+## Integration with R007
+
+R008 extends R007 for tool operations:
+
+```
+┌─ Agent: creator (manager)
+├─ Model: sonnet
+└─ Task: Creating new agent
+
+[creator][sonnet] → Write: agents/sw-engineer/new-agent/AGENT.md
+[creator][sonnet] → Write: agents/sw-engineer/new-agent/index.yaml
+[creator][sonnet] → Edit: agents/index.yaml
+
+[Done] Agent created successfully
+```
+
+## Benefits
+
+1. **Debugging**: Know which agent performed which operation
+2. **Transparency**: User sees all agent activities
+3. **Audit Trail**: Track file changes by agent
+4. **Error Attribution**: Identify which agent caused issues
+
+## Implementation
+
+Agents should:
+1. Identify themselves before tool operations
+2. Include tool name and target
+3. Use consistent format across all tools
+4. Group related operations together

package/templates/agents/manager/creator/refs/rules/SHOULD-ecomode.md

@@ -0,0 +1,145 @@
+# [SHOULD] Ecomode Rules
+
+> **Priority**: SHOULD - Recommended for token efficiency
+> **ID**: R013
+
+## Purpose
+
+Reduce token usage and improve efficiency during batch operations and parallel execution.
+
+## Activation Conditions
+
+Ecomode is automatically activated when:
+
+```yaml
+conditions:
+  - 4+ parallel tasks spawned
+  - Batch operations requested
+  - Near compaction threshold (80%+ context usage)
+  - Explicit user request: "ecomode on"
+```
+
+## Behaviors
+
+### 1. Compact Output Format
+
+When ecomode is active, parallel agents return:
+
+```yaml
+format:
+  - status: success | failed | partial
+  - summary: 1-2 sentences max
+  - key_data: essential results only
+
+skip:
+  - Intermediate steps
+  - Verbose explanations
+  - Repeated context
+  - Full file contents (use paths instead)
+```
+
+### 2. Aggregation Format
+
+Secretary aggregates results in compact format:
+
+```
+[Batch Complete] {n}/{total}
+├── {agent}: {icon} {summary}
+├── {agent}: {icon} {summary}
+└── {agent}: {icon} {summary}
+```
+
+Icons:
+- ✓ success
+- ✗ failed
+- ⚠ partial/warning
+
+### 3. Result Compression
+
+```yaml
+compress:
+  - File lists → count only unless < 5 files
+  - Error traces → first and last 3 lines
+  - Code snippets → path:line reference only
+  - Long outputs → truncate with "[truncated]"
+```
+
+## Secretary Ecomode Config
+
+```yaml
+ecomode:
+  enabled: true
+  threshold: 4           # Activate when >= 4 parallel tasks
+  result_format: summary # summary | full
+  max_result_length: 200 # characters per result
+```
+
+## Examples
+
+### Normal Mode Output
+
+```
+┌─ Agent: golang-expert (sw-engineer)
+├─ Skill: go-best-practices
+└─ Task: Reviewing src/main.go
+
+Reading file...
+[golang-expert → Read] src/main.go
+
+Analyzing code structure...
+Found 3 functions, 2 structs, 1 interface.
+
+Checking naming conventions...
+Function 'GetUser' follows Go naming conventions.
+Function 'processData' follows Go naming conventions.
+Function 'handle_error' VIOLATION: uses snake_case instead of camelCase.
+
+[...]
+
+Review Complete:
+- 1 naming violation found
+- 2 potential improvements suggested
+- Overall code quality: Good
+```
+
+### Ecomode Output
+
+```
+[golang-expert] ✓ src/main.go reviewed: 1 naming issue (handle_error → handleError)
+```
+
+## Implementation Notes
+
+### For Orchestrator (Secretary)
+
+```yaml
+responsibilities:
+  - Detect ecomode activation conditions
+  - Instruct spawned agents to use compact format
+  - Aggregate results in batch format
+  - Track token savings
+```
+
+### For Worker Agents
+
+```yaml
+when_ecomode_active:
+  - Return status + summary only
+  - Skip intermediate progress updates
+  - Use references instead of full content
+  - Compress error messages
+```
+
+## Override
+
+User can disable ecomode:
+- "ecomode off"
+- "verbose mode"
+- "show full details"
+
+## Benefits
+
+1. **Token Efficiency**: 60-80% reduction in batch operations
+2. **Faster Response**: Less output processing
+3. **Better Overview**: Aggregated results at a glance
+4. **Context Preservation**: More room for actual work

package/templates/agents/manager/creator/refs/rules/SHOULD-error-handling.md

@@ -0,0 +1,102 @@
+# [SHOULD] Error Handling Rules
+
+> **Priority**: SHOULD - Strongly recommended
+> **Goal**: Safe failure, fast recovery
+
+## Error Classification
+
+### Level 1: Warning
+```
+Symptom: Task completes but needs attention
+Response: Output warning, continue
+Example:
+  - Non-recommended pattern found
+  - Potential performance issue
+  - Better alternative exists
+```
+
+### Level 2: Error
+```
+Symptom: Current task fails, others possible
+Response: Stop task, report cause, suggest alternative
+Example:
+  - File not found
+  - Insufficient permission
+  - Format error
+```
+
+### Level 3: Critical
+```
+Symptom: Cannot proceed at all
+Response: Stop all, preserve state, report immediately
+Example:
+  - System resource exhausted
+  - Missing required dependency
+  - Security violation detected
+```
+
+## Error Report Format
+
+```
+[Error] {error type}
+
+Location: {file:line or task name}
+Cause: {specific cause}
+Impact: {effect of this error}
+
+Attempted:
+1. {attempt 1} → Failed
+2. {attempt 2} → Failed
+
+Recommended:
+- {action 1}
+- {action 2}
+```
+
+## Recovery Strategy
+
+### Retryable Errors
+```
+1. Retry up to 3 times
+2. Wait between retries (1s, 2s, 4s)
+3. Report to user after 3 failures
+```
+
+### Non-recoverable Errors
+```
+1. Save current state
+2. Rollback changes (if possible)
+3. Detailed error report
+4. Wait for user instruction
+```
+
+## Preventive Validation
+
+### Before Action
+```
+□ Target file/path exists
+□ Required permissions available
+□ Dependencies met
+□ Sufficient resources
+```
+
+### After Action
+```
+□ Expected result matches actual
+□ File integrity verified
+□ No side effects
+```
+
+## Error Logging
+
+```yaml
+error_log:
+  timestamp: "2025-01-22T10:30:00Z"
+  level: error
+  code: FILE_NOT_FOUND
+  message: "Config file not found"
+  context:
+    file: "/path/to/config.yaml"
+    action: "read"
+  resolution: "Using default config"
+```