@namch/agent-assistant 1.1.1 โ 1.2.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +35 -1
- package/README.md +7 -2
- package/agents/backend-engineer.md +1 -1
- package/agents/brainstormer.md +1 -1
- package/agents/business-analyst.md +1 -1
- package/agents/database-architect.md +1 -1
- package/agents/debugger.md +1 -1
- package/agents/designer.md +1 -1
- package/agents/devops-engineer.md +1 -1
- package/agents/docs-manager.md +1 -1
- package/agents/frontend-engineer.md +1 -1
- package/agents/game-engineer.md +1 -1
- package/agents/mobile-engineer.md +1 -1
- package/agents/performance-engineer.md +1 -1
- package/agents/planner.md +1 -1
- package/agents/project-manager.md +1 -1
- package/agents/reporter.md +1 -1
- package/agents/researcher.md +1 -1
- package/agents/reviewer.md +1 -1
- package/agents/scouter.md +1 -1
- package/agents/security-engineer.md +1 -1
- package/agents/teams/backend-team/executor.md +188 -0
- package/agents/teams/backend-team/reviewer.md +328 -0
- package/agents/teams/backend-team/techlead.md +166 -0
- package/agents/teams/database-team/executor.md +189 -0
- package/agents/teams/database-team/reviewer.md +333 -0
- package/agents/teams/database-team/techlead.md +168 -0
- package/agents/teams/debug-team/executor.md +195 -0
- package/agents/teams/debug-team/reviewer.md +324 -0
- package/agents/teams/debug-team/techlead.md +173 -0
- package/agents/teams/design-team/executor.md +196 -0
- package/agents/teams/design-team/reviewer.md +405 -0
- package/agents/teams/design-team/techlead.md +170 -0
- package/agents/teams/devops-team/executor.md +199 -0
- package/agents/teams/devops-team/reviewer.md +332 -0
- package/agents/teams/devops-team/techlead.md +168 -0
- package/agents/teams/docs-team/executor.md +196 -0
- package/agents/teams/docs-team/reviewer.md +331 -0
- package/agents/teams/docs-team/techlead.md +168 -0
- package/agents/teams/frontend-team/executor.md +190 -0
- package/agents/teams/frontend-team/reviewer.md +333 -0
- package/agents/teams/frontend-team/techlead.md +169 -0
- package/agents/teams/fullstack-team/executor.md +290 -0
- package/agents/teams/fullstack-team/reviewer.md +365 -0
- package/agents/teams/fullstack-team/techlead.md +254 -0
- package/agents/teams/game-team/executor.md +193 -0
- package/agents/teams/game-team/reviewer.md +331 -0
- package/agents/teams/game-team/techlead.md +167 -0
- package/agents/teams/mobile-team/executor.md +192 -0
- package/agents/teams/mobile-team/reviewer.md +328 -0
- package/agents/teams/mobile-team/techlead.md +168 -0
- package/agents/teams/performance-team/executor.md +192 -0
- package/agents/teams/performance-team/reviewer.md +322 -0
- package/agents/teams/performance-team/techlead.md +175 -0
- package/agents/teams/planning-team/executor.md +197 -0
- package/agents/teams/planning-team/reviewer.md +279 -0
- package/agents/teams/planning-team/techlead.md +169 -0
- package/agents/teams/project-team/executor.md +190 -0
- package/agents/teams/project-team/reviewer.md +328 -0
- package/agents/teams/project-team/techlead.md +168 -0
- package/agents/teams/qa-team/executor.md +198 -0
- package/agents/teams/qa-team/reviewer.md +271 -0
- package/agents/teams/qa-team/techlead.md +175 -0
- package/agents/teams/report-team/executor.md +195 -0
- package/agents/teams/report-team/reviewer.md +328 -0
- package/agents/teams/report-team/techlead.md +168 -0
- package/agents/teams/research-team/executor.md +200 -0
- package/agents/teams/research-team/reviewer.md +272 -0
- package/agents/teams/research-team/techlead.md +168 -0
- package/agents/teams/security-team/executor.md +193 -0
- package/agents/teams/security-team/reviewer.md +338 -0
- package/agents/teams/security-team/techlead.md +178 -0
- package/agents/tech-lead.md +1 -1
- package/agents/tester.md +1 -1
- package/cli/README.md +19 -2
- package/cli/install.js +282 -7
- package/cli/install.test.js.example +1 -1
- package/code-assistants/codex-assistant/CODEX.md +235 -0
- package/code-assistants/codex-assistant/agents/backend-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/brainstormer.toml +26 -0
- package/code-assistants/codex-assistant/agents/business-analyst.toml +26 -0
- package/code-assistants/codex-assistant/agents/database-architect.toml +26 -0
- package/code-assistants/codex-assistant/agents/debugger.toml +26 -0
- package/code-assistants/codex-assistant/agents/designer.toml +26 -0
- package/code-assistants/codex-assistant/agents/devops-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/docs-manager.toml +26 -0
- package/code-assistants/codex-assistant/agents/frontend-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/game-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/mobile-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/performance-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/planner.toml +26 -0
- package/code-assistants/codex-assistant/agents/project-manager.toml +26 -0
- package/code-assistants/codex-assistant/agents/reporter.toml +26 -0
- package/code-assistants/codex-assistant/agents/researcher.toml +26 -0
- package/code-assistants/codex-assistant/agents/reviewer.toml +26 -0
- package/code-assistants/codex-assistant/agents/scouter.toml +26 -0
- package/code-assistants/codex-assistant/agents/security-engineer.toml +26 -0
- package/code-assistants/codex-assistant/agents/tech-lead.toml +26 -0
- package/code-assistants/codex-assistant/agents/tester.toml +26 -0
- package/code-assistants/codex-assistant/config.toml +109 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-ask-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-auto/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-auto/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-brainstorm-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-code-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-cook-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-debug-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-check/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-check/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-preview/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-preview/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-production/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-production/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-rollback/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-deploy-rollback/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-design-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-audit/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-audit/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-business/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-business/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-core/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-docs-core/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-fix-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-plan-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-report-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-review-team/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-fast/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-fast/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-focus/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-focus/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-hard/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-hard/agents/openai.yaml +4 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-team/SKILL.md +18 -0
- package/code-assistants/codex-assistant/skills/agent-assistant-test-team/agents/openai.yaml +4 -0
- package/code-assistants/copilot-assistant/agent-assistant.agent.md +0 -25
- package/commands/brainstorm/team.md +295 -0
- package/commands/brainstorm.md +5 -0
- package/commands/code/team.md +456 -0
- package/commands/code.md +5 -0
- package/commands/cook/team.md +609 -0
- package/commands/cook.md +5 -0
- package/commands/debug/team.md +396 -0
- package/commands/debug.md +5 -0
- package/commands/design/team.md +567 -0
- package/commands/design.md +5 -0
- package/commands/fix/team.md +596 -0
- package/commands/fix.md +5 -0
- package/commands/plan/team.md +358 -0
- package/commands/plan.md +5 -0
- package/commands/report/team.md +502 -0
- package/commands/report.md +5 -0
- package/commands/review/team.md +353 -0
- package/commands/review.md +5 -0
- package/commands/test/team.md +303 -0
- package/commands/test.md +5 -0
- package/documents/SMART-SKILL-ORCHESTRATION-BLUEPRINT.md +4 -2
- package/documents/business/business-prd.md +2 -1
- package/documents/business/business-workflows.md +2 -2
- package/documents/knowledge-architecture.md +13 -12
- package/documents/knowledge-domain.md +2 -2
- package/documents/knowledge-overview.md +2 -2
- package/documents/knowledge-source-base.md +4 -0
- package/package.json +5 -2
- package/rules/AGENTS.md +55 -0
- package/rules/CORE.md +1 -0
- package/rules/PHASES.md +58 -0
- package/rules/TEAMS.md +530 -0
package/rules/PHASES.md
CHANGED
|
@@ -49,6 +49,64 @@
|
|
|
49
49
|
|
|
50
50
|
---
|
|
51
51
|
|
|
52
|
+
## ๐บ GOLDEN TRIANGLE PHASE OUTPUT FORMAT (`:team` variant only)
|
|
53
|
+
|
|
54
|
+
> **LOAD**: `TEAMS.md` for full Golden Triangle protocol.
|
|
55
|
+
> When a workflow uses `:team` variant, replace the standard phase format above with this format.
|
|
56
|
+
> Each phase spawns exactly **3 agents**: Tech Lead + Executor + Reviewer.
|
|
57
|
+
|
|
58
|
+
### Emit progressively:
|
|
59
|
+
|
|
60
|
+
```markdown
|
|
61
|
+
## ๐ญ Phase {N}: {name} โ ๐บ GOLDEN TRIANGLE
|
|
62
|
+
|
|
63
|
+
### ๐บ Triangle Assignment
|
|
64
|
+
| Role | Agent | Mission |
|
|
65
|
+
|------|-------|---------|
|
|
66
|
+
| Tech Lead | `{agent}` | {coordination mission} |
|
|
67
|
+
| Executor | `{agent}` | {implementation mission} |
|
|
68
|
+
| Reviewer | `{agent}` | {review mission} |
|
|
69
|
+
|
|
70
|
+
### ๐ Shared Task List
|
|
71
|
+
| # | Task | Owner | Status | Round |
|
|
72
|
+
|---|------|-------|--------|-------|
|
|
73
|
+
| 1 | {task} | `{executor}` | โ
/๐/โณ | 1/3 |
|
|
74
|
+
|
|
75
|
+
### ๐ฌ Mailbox Exchanges (key entries from ./reports/MAILBOX-{date}.md)
|
|
76
|
+
| # | From โ To | Type | Summary |
|
|
77
|
+
|---|-----------|------|---------|
|
|
78
|
+
| 1 | Tech Lead โ Executor | TASK_ASSIGNMENT | {summary} |
|
|
79
|
+
| 2 | Executor โ Reviewer | SUBMISSION | {summary} |
|
|
80
|
+
| 3 | Reviewer โ Executor | REVIEW (PASS/FAIL) | {summary} |
|
|
81
|
+
| 4 | Executor โ Reviewer | DEFENSE/RESUBMISSION | {summary} |
|
|
82
|
+
| 5 | Tech Lead โ ALL | DECISION | {summary} |
|
|
83
|
+
|
|
84
|
+
### ๐ Debate Summary (if any)
|
|
85
|
+
- **Rounds used**: {N}/3
|
|
86
|
+
- **Key disputes**: {brief}
|
|
87
|
+
- **Resolution**: {PASS / Tech Lead arbitration}
|
|
88
|
+
|
|
89
|
+
### โ
CONSENSUS: {TechLead} โ | {Executor} โ | {Reviewer} โ
|
|
90
|
+
|
|
91
|
+
### Exit Criteria
|
|
92
|
+
- [x] {criterion_1}
|
|
93
|
+
- [x] {criterion_2}
|
|
94
|
+
|
|
95
|
+
### โ
Phase {N} complete โ Golden Triangle released
|
|
96
|
+
**Deliverable**: {summary}
|
|
97
|
+
```
|
|
98
|
+
|
|
99
|
+
**Rules**:
|
|
100
|
+
- Team phases use Golden Triangle format; non-team phases use standard format
|
|
101
|
+
- ALWAYS exactly 3 agents per phase (Tech Lead + Executor + Reviewer)
|
|
102
|
+
- Tech Lead decomposes โ Executor implements โ Reviewer critiques โ Debate โ Consensus
|
|
103
|
+
- Mailbox (`./reports/MAILBOX-{date}.md`) captures ALL inter-agent communication
|
|
104
|
+
- Maximum 3 debate rounds per task โ then Tech Lead arbitrates
|
|
105
|
+
- Output released ONLY with explicit consensus stamp: `โ
CONSENSUS: X โ | Y โ | Z โ`
|
|
106
|
+
- NO phase can complete without all 3 agents confirming
|
|
107
|
+
|
|
108
|
+
---
|
|
109
|
+
|
|
52
110
|
## PHASE EXECUTION RULES
|
|
53
111
|
|
|
54
112
|
### One Phase at a Time (No Batching)
|
package/rules/TEAMS.md
ADDED
|
@@ -0,0 +1,530 @@
|
|
|
1
|
+
# ๐บ TEAMS โ Golden Triangle Architecture
|
|
2
|
+
|
|
3
|
+
> **LOAD**: When `:team` variant is invoked | **PURPOSE**: 3-agent adversarial collaboration protocol
|
|
4
|
+
> **VERSION**: 2.0 โ Golden Triangle | **SUPERSEDES**: Team Lead + Teammates model
|
|
5
|
+
|
|
6
|
+
---
|
|
7
|
+
|
|
8
|
+
## CORE PRINCIPLE
|
|
9
|
+
|
|
10
|
+
Every team phase spawns exactly **3 agent roles** โ no more, no less. Quality emerges from structured tension between an Executor who builds and a Reviewer who challenges, orchestrated by a Tech Lead who arbitrates.
|
|
11
|
+
|
|
12
|
+
```
|
|
13
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
14
|
+
โ ORCHESTRATOR โ
|
|
15
|
+
โ โโโ invokes Golden Triangle for Phase N โ
|
|
16
|
+
โ โ
|
|
17
|
+
โ โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ โ
|
|
18
|
+
โ โ ๐บ TECH LEAD โ โ
|
|
19
|
+
โ โ Decomposes ยท Arbitrates โ โ
|
|
20
|
+
โ โ Synthesizes ยท FINAL AUTHORITY โ โ
|
|
21
|
+
โ โโโโโโโโโโโโฌโโโโโโโโโโโฌโโโโโโโโโโโโ โ
|
|
22
|
+
โ โ โ โ
|
|
23
|
+
โ โโโโโโโโโผโโโ โโโโโผโโโโโโโโโ โ
|
|
24
|
+
โ โ EXECUTOR โ โ REVIEWER โ โ
|
|
25
|
+
โ โ Builds โโโโค Challenges โ โ
|
|
26
|
+
โ โ Defends โโโโบโ Validates โ โ
|
|
27
|
+
โ โโโโโโโโโโโโ โโโโโโโโโโโโโ โ
|
|
28
|
+
โ โฒ โฒ โ
|
|
29
|
+
โ โโโโโ ๐ฌ โโโ โ
|
|
30
|
+
โ MAILBOX โ
|
|
31
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
32
|
+
```
|
|
33
|
+
|
|
34
|
+
---
|
|
35
|
+
|
|
36
|
+
## THE THREE ROLES
|
|
37
|
+
|
|
38
|
+
### 1. Tech Lead (`techlead.md`)
|
|
39
|
+
|
|
40
|
+
| Attribute | Description |
|
|
41
|
+
|-----------|-------------|
|
|
42
|
+
| **Function** | Task decomposer, work coordinator, dispute arbitrator, final output synthesizer |
|
|
43
|
+
| **Authority** | FINAL on all decisions โ overrides Executor and Reviewer when consensus fails |
|
|
44
|
+
| **Actions** | Produces Shared Task List, reads all Mailbox exchanges, issues binding DECISION |
|
|
45
|
+
| **Personality** | Pragmatic, decisive, plan-focused. Minimizes debate rounds. Ships quality work. |
|
|
46
|
+
|
|
47
|
+
### 2. Executor (`executor.md`)
|
|
48
|
+
|
|
49
|
+
| Attribute | Description |
|
|
50
|
+
|-----------|-------------|
|
|
51
|
+
| **Function** | Direct implementer โ builds, codes, and creates the actual deliverable |
|
|
52
|
+
| **Authority** | Owns implementation decisions. Can and MUST defend work when Reviewer is wrong. |
|
|
53
|
+
| **Actions** | Implements tasks, posts SUBMISSION to Mailbox, posts DEFENSE with evidence when challenged |
|
|
54
|
+
| **Personality** | Builder mindset. Pragmatic, evidence-driven. Does NOT blindly accept all feedback. Pushes back with technical proof when Reviewer's critique is incorrect or over-engineered. |
|
|
55
|
+
|
|
56
|
+
### 3. Reviewer (`reviewer.md`)
|
|
57
|
+
|
|
58
|
+
| Attribute | Description |
|
|
59
|
+
|-----------|-------------|
|
|
60
|
+
| **Function** | Quality gatekeeper with Devil's Advocate mindset |
|
|
61
|
+
| **Authority** | Can FAIL submissions and demand fixes. Can escalate unresolved disputes to Tech Lead. |
|
|
62
|
+
| **Actions** | Reviews submissions, posts REVIEW (PASS/FAIL) to Mailbox, re-checks fixes and defenses |
|
|
63
|
+
| **Personality** | Skeptical, thorough, demanding. Assumes defects exist until proven otherwise. Checks security, performance, correctness, edge cases. Does NOT rubber-stamp. |
|
|
64
|
+
|
|
65
|
+
---
|
|
66
|
+
|
|
67
|
+
## THE DEBATE MECHANISM
|
|
68
|
+
|
|
69
|
+
No 4th agent is needed. Adversarial tension is **hardcoded** into Executor and Reviewer personalities.
|
|
70
|
+
|
|
71
|
+
```
|
|
72
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
73
|
+
โ DEBATE FLOW โ
|
|
74
|
+
โ โ
|
|
75
|
+
โ Executor implements โโโบ Reviewer critiques โ
|
|
76
|
+
โ โ โ โ
|
|
77
|
+
โ โ โโโโโโโโโโโโโโโโโโโโโโ โ
|
|
78
|
+
โ โ โ โ
|
|
79
|
+
โ โผ โผ โ
|
|
80
|
+
โ Executor DEFENDS (with evidence) โโโ OR โโโ FIXES โ
|
|
81
|
+
โ โ โ
|
|
82
|
+
โ โผ โ
|
|
83
|
+
โ Reviewer re-checks โโโบ PASS โโโบ Consensus โ
|
|
84
|
+
โ โ โ
|
|
85
|
+
โ โโโโบ FAIL โโโบ Loop (max 3 rounds) โ
|
|
86
|
+
โ โ โ
|
|
87
|
+
โ โผ โ
|
|
88
|
+
โ Tech Lead ARBITRATES (binding) โ
|
|
89
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
90
|
+
```
|
|
91
|
+
|
|
92
|
+
**Max debate rounds**: 3. After round 3 without agreement, Tech Lead reads all Mailbox exchanges and makes a **binding decision**. No further debate.
|
|
93
|
+
|
|
94
|
+
**Defense rules**:
|
|
95
|
+
- Executor MUST defend with **technical evidence** (benchmarks, specs, code references)
|
|
96
|
+
- "I disagree" without proof = automatic FAIL, Reviewer wins
|
|
97
|
+
- Reviewer MUST consider valid evidence. Rejecting proven-correct work = escalation to Tech Lead
|
|
98
|
+
- Tech Lead evaluates **evidence quality**, not seniority or role
|
|
99
|
+
|
|
100
|
+
---
|
|
101
|
+
|
|
102
|
+
## COMMUNICATION VIA SHARED FILES
|
|
103
|
+
|
|
104
|
+
All agent communication flows through two shared artifacts:
|
|
105
|
+
|
|
106
|
+
| Artifact | Owner | Purpose |
|
|
107
|
+
|----------|-------|---------|
|
|
108
|
+
| **Shared Task List** | Tech Lead | State management for the phase's tasks โ assignments, status, priorities |
|
|
109
|
+
| **Mailbox** | All agents (append-only) | `./reports/MAILBOX-{date}.md` โ communication log for submissions, reviews, defenses, decisions |
|
|
110
|
+
|
|
111
|
+
**Rules**:
|
|
112
|
+
- Mailbox is **append-only** โ no agent may edit or delete prior exchanges
|
|
113
|
+
- All agents read the full Mailbox to maintain shared context
|
|
114
|
+
- Tech Lead manages the Shared Task List; Executor and Reviewer read it
|
|
115
|
+
- One Mailbox per phase execution, timestamped by date
|
|
116
|
+
|
|
117
|
+
---
|
|
118
|
+
|
|
119
|
+
## COMMUNICATION LOOP (Per Phase)
|
|
120
|
+
|
|
121
|
+
```
|
|
122
|
+
1. Tech Lead decomposes task
|
|
123
|
+
โ Publishes Shared Task List with assignments
|
|
124
|
+
โ Posts TASK_ASSIGNMENT to Mailbox
|
|
125
|
+
|
|
126
|
+
2. Executor receives task
|
|
127
|
+
โ Implements deliverable
|
|
128
|
+
โ Posts SUBMISSION to Mailbox: description, files changed, approach
|
|
129
|
+
|
|
130
|
+
3. Reviewer reads submission
|
|
131
|
+
โ Reviews for quality, security, performance, correctness
|
|
132
|
+
โ Posts REVIEW to Mailbox: findings, PASS or FAIL
|
|
133
|
+
|
|
134
|
+
4. IF FAIL:
|
|
135
|
+
a. Executor reads review findings
|
|
136
|
+
b. For each finding, Executor either:
|
|
137
|
+
โ FIXES the issue and resubmits, OR
|
|
138
|
+
โ DEFENDS with technical evidence via Mailbox
|
|
139
|
+
c. IF defense posted:
|
|
140
|
+
โ Reviewer considers evidence โ accepts defense OR rejects with counter-evidence
|
|
141
|
+
โ IF unresolved after re-check โ Reviewer escalates to Tech Lead
|
|
142
|
+
d. Return to step 3 (max 3 rounds total)
|
|
143
|
+
|
|
144
|
+
5. IF PASS (or max 3 rounds reached):
|
|
145
|
+
โ Tech Lead reads ALL Mailbox exchanges
|
|
146
|
+
โ Evaluates evidence from both sides if disputes exist
|
|
147
|
+
โ Makes final synthesis decision
|
|
148
|
+
โ Posts DECISION to Mailbox with approved output
|
|
149
|
+
|
|
150
|
+
6. Phase output released with CONSENSUS stamp
|
|
151
|
+
โ All three agents sign off
|
|
152
|
+
โ Deliverable is final
|
|
153
|
+
```
|
|
154
|
+
|
|
155
|
+
---
|
|
156
|
+
|
|
157
|
+
## CONSENSUS PROTOCOL
|
|
158
|
+
|
|
159
|
+
Output is released **ONLY** when one of these conditions is met:
|
|
160
|
+
|
|
161
|
+
| Condition | Trigger |
|
|
162
|
+
|-----------|---------|
|
|
163
|
+
| **Clean pass** | Reviewer explicitly APPROVED Executor's submission (no disputes) |
|
|
164
|
+
| **Resolved pass** | Reviewer APPROVED after Executor fixed issues or defended successfully |
|
|
165
|
+
| **Arbitrated pass** | Tech Lead overrode after max 3 rounds โ documents reasoning |
|
|
166
|
+
|
|
167
|
+
**Consensus stamp format**:
|
|
168
|
+
```
|
|
169
|
+
โ
CONSENSUS: TechLead โ | Executor โ | Reviewer โ
|
|
170
|
+
```
|
|
171
|
+
|
|
172
|
+
**Without this stamp, no phase output is released.** If any agent has not signed off, Tech Lead must resolve the gap before proceeding.
|
|
173
|
+
|
|
174
|
+
---
|
|
175
|
+
|
|
176
|
+
## TEAM ROSTER (Agent-to-Role Mapping per Domain)
|
|
177
|
+
|
|
178
|
+
| Team Domain | Tech Lead Agent | Executor Agent | Reviewer Focus |
|
|
179
|
+
|-------------|----------------|----------------|----------------|
|
|
180
|
+
| `backend` | `tech-lead` | `backend-engineer` | `reviewer` โ security + performance |
|
|
181
|
+
| `frontend` | `tech-lead` | `frontend-engineer` | `reviewer` โ design + performance |
|
|
182
|
+
| `fullstack` | `tech-lead` | `backend-engineer` + `frontend-engineer` | `reviewer` โ security + performance |
|
|
183
|
+
| `database` | `tech-lead` | `database-architect` | `reviewer` โ security + performance |
|
|
184
|
+
| `research` | `researcher` | `scouter` | `brainstormer` โ critical evaluator |
|
|
185
|
+
| `planning` | `planner` | `researcher` | `tech-lead` โ feasibility critic |
|
|
186
|
+
| `qa` | `tester` | `tester` | `security-engineer` + `performance-engineer` |
|
|
187
|
+
| `design` | `designer` | `frontend-engineer` | `reviewer` โ UX + accessibility |
|
|
188
|
+
| `debug` | `debugger` | `backend-engineer` | `reviewer` โ root-cause validator |
|
|
189
|
+
| `devops` | `devops-engineer` | `backend-engineer` | `security-engineer` |
|
|
190
|
+
| `security` | `security-engineer` | `backend-engineer` | `reviewer` โ pen-test mindset |
|
|
191
|
+
| `game` | `tech-lead` | `game-engineer` | `reviewer` โ game architecture + performance (60fps) |
|
|
192
|
+
| `mobile` | `tech-lead` | `mobile-engineer` | `reviewer` โ UX + platform compliance + performance |
|
|
193
|
+
| `performance` | `performance-engineer` | `backend-engineer` | `reviewer` โ measurement validation + regression |
|
|
194
|
+
| `docs` | `docs-manager` | `researcher` | `reviewer` โ accuracy + completeness + clarity |
|
|
195
|
+
| `project` | `project-manager` | `business-analyst` | `tech-lead` โ technical feasibility |
|
|
196
|
+
| `report` | `reporter` | `scouter` | `reviewer` โ data accuracy + insight validity |
|
|
197
|
+
|
|
198
|
+
### Team Selection
|
|
199
|
+
|
|
200
|
+
```
|
|
201
|
+
IF workflow specifies a team domain โ USE that domain's row from roster
|
|
202
|
+
ELSE IF phase maps to a domain โ USE default mapping
|
|
203
|
+
ELSE โ HALT, ask Orchestrator for team assignment
|
|
204
|
+
```
|
|
205
|
+
|
|
206
|
+
### Fullstack Special Case
|
|
207
|
+
|
|
208
|
+
The `fullstack` domain spawns **two Executors** (backend + frontend). They share the Mailbox and take turns submitting. Reviewer reviews each submission independently. Tech Lead coordinates integration between the two.
|
|
209
|
+
|
|
210
|
+
---
|
|
211
|
+
|
|
212
|
+
## TIERED EXECUTION FOR GOLDEN TRIANGLE
|
|
213
|
+
|
|
214
|
+
### TIER 1 โ Sub-agent Spawn (preferred)
|
|
215
|
+
|
|
216
|
+
```
|
|
217
|
+
1. Orchestrator spawns Tech Lead as sub-agent
|
|
218
|
+
2. Tech Lead spawns Executor as sub-agent
|
|
219
|
+
3. Tech Lead spawns Reviewer as sub-agent
|
|
220
|
+
4. Communication flows via Mailbox file (./reports/MAILBOX-{date}.md)
|
|
221
|
+
5. Debate loop runs until consensus or max rounds
|
|
222
|
+
6. Tech Lead collects final output and returns to Orchestrator
|
|
223
|
+
```
|
|
224
|
+
|
|
225
|
+
### TIER 2 โ Sequential Embodiment (fallback)
|
|
226
|
+
|
|
227
|
+
```
|
|
228
|
+
1. EMBODY Tech Lead โ decompose task โ produce Shared Task List โ post TASK_ASSIGNMENT
|
|
229
|
+
2. EMBODY Executor โ implement tasks โ post SUBMISSION to Mailbox
|
|
230
|
+
3. EMBODY Reviewer โ review submission โ post REVIEW to Mailbox (PASS/FAIL)
|
|
231
|
+
4. IF FAIL:
|
|
232
|
+
a. Re-EMBODY Executor โ read findings โ FIX or DEFEND โ post to Mailbox
|
|
233
|
+
b. Re-EMBODY Reviewer โ re-check โ post updated REVIEW
|
|
234
|
+
c. Repeat steps 4a-4b (max 3 total rounds)
|
|
235
|
+
5. EMBODY Tech Lead โ read all Mailbox exchanges โ arbitrate if needed โ synthesize
|
|
236
|
+
6. Post DECISION with consensus stamp
|
|
237
|
+
```
|
|
238
|
+
|
|
239
|
+
```
|
|
240
|
+
โ FORBIDDEN: Using TIER 2 when runSubagent is available
|
|
241
|
+
โ FORBIDDEN: Skipping the Reviewer step (all work MUST be reviewed)
|
|
242
|
+
โ FORBIDDEN: Executor and Reviewer directly modifying each other's outputs
|
|
243
|
+
โ
REQUIRED: Attempt TIER 1 first, log reason if falling back to TIER 2
|
|
244
|
+
โ
REQUIRED: Mailbox used for ALL inter-agent communication in both tiers
|
|
245
|
+
```
|
|
246
|
+
|
|
247
|
+
---
|
|
248
|
+
|
|
249
|
+
## SHARED TASK LIST FORMAT
|
|
250
|
+
|
|
251
|
+
**โ Tech Lead MUST produce this before any Executor work begins.**
|
|
252
|
+
|
|
253
|
+
```markdown
|
|
254
|
+
## ๐ Shared Task List โ {Phase Name}
|
|
255
|
+
|
|
256
|
+
| ID | Task | Assigned To | Status | Priority | Round |
|
|
257
|
+
|----|------|-------------|--------|----------|-------|
|
|
258
|
+
| T1 | {task description} | `executor` | โณ | H | 1 |
|
|
259
|
+
| T2 | {task description} | `executor` | โณ | M | 1 |
|
|
260
|
+
```
|
|
261
|
+
|
|
262
|
+
**Status icons**:
|
|
263
|
+
| Icon | Meaning |
|
|
264
|
+
|------|---------|
|
|
265
|
+
| โณ | Pending โ not started |
|
|
266
|
+
| ๐ | In Progress โ Executor working |
|
|
267
|
+
| โ
| Complete + Approved โ passed review |
|
|
268
|
+
| โ | Blocked โ dependency or unresolved issue |
|
|
269
|
+
| ๐ | Revision Needed โ failed review, requires fix |
|
|
270
|
+
|
|
271
|
+
**Rules**:
|
|
272
|
+
- All implementation tasks are assigned to `executor` (there is only one Executor per triangle)
|
|
273
|
+
- Tech Lead updates status as the debate loop progresses
|
|
274
|
+
- `Round` tracks which debate round the task is on (1, 2, or 3)
|
|
275
|
+
- A task is โ
only when Reviewer has approved OR Tech Lead has arbitrated
|
|
276
|
+
|
|
277
|
+
---
|
|
278
|
+
|
|
279
|
+
## MAILBOX FILE FORMAT
|
|
280
|
+
|
|
281
|
+
**File**: `./reports/MAILBOX-{date}.md`
|
|
282
|
+
|
|
283
|
+
```markdown
|
|
284
|
+
# ๐ฌ MAILBOX โ {Phase Name} โ {date}
|
|
285
|
+
|
|
286
|
+
## Exchange #1
|
|
287
|
+
| From | To | Type | Timestamp |
|
|
288
|
+
|------|----|------|-----------|
|
|
289
|
+
| `tech-lead` | `all` | TASK_ASSIGNMENT | {time} |
|
|
290
|
+
|
|
291
|
+
**Content:**
|
|
292
|
+
[Shared Task List with assignments and context]
|
|
293
|
+
|
|
294
|
+
---
|
|
295
|
+
|
|
296
|
+
## Exchange #2
|
|
297
|
+
| From | To | Type | Timestamp |
|
|
298
|
+
|------|----|------|-----------|
|
|
299
|
+
| `executor` | `reviewer` | SUBMISSION | {time} |
|
|
300
|
+
|
|
301
|
+
**Content:**
|
|
302
|
+
[Description of what was implemented, files changed, approach taken]
|
|
303
|
+
|
|
304
|
+
---
|
|
305
|
+
|
|
306
|
+
## Exchange #3
|
|
307
|
+
| From | To | Type | Timestamp |
|
|
308
|
+
|------|----|------|-----------|
|
|
309
|
+
| `reviewer` | `executor` | REVIEW | {time} |
|
|
310
|
+
|
|
311
|
+
**Status:** โ FAIL
|
|
312
|
+
**Findings:**
|
|
313
|
+
- [Issue 1: severity HIGH โ description โ required fix]
|
|
314
|
+
- [Issue 2: severity MEDIUM โ description โ required fix]
|
|
315
|
+
|
|
316
|
+
---
|
|
317
|
+
|
|
318
|
+
## Exchange #4
|
|
319
|
+
| From | To | Type | Timestamp |
|
|
320
|
+
|------|----|------|-----------|
|
|
321
|
+
| `executor` | `reviewer` | DEFENSE | {time} |
|
|
322
|
+
|
|
323
|
+
**Content:**
|
|
324
|
+
[Fix applied for Issue 1 โ files changed, approach]
|
|
325
|
+
[Technical evidence defending approach for Issue 2 โ benchmarks, spec references, rationale]
|
|
326
|
+
|
|
327
|
+
---
|
|
328
|
+
|
|
329
|
+
## Exchange #5
|
|
330
|
+
| From | To | Type | Timestamp |
|
|
331
|
+
|------|----|------|-----------|
|
|
332
|
+
| `reviewer` | `tech-lead` | APPROVAL | {time} |
|
|
333
|
+
|
|
334
|
+
**Status:** โ
PASS
|
|
335
|
+
Executor addressed Issue 1. Defense for Issue 2 accepted โ evidence verified.
|
|
336
|
+
|
|
337
|
+
---
|
|
338
|
+
|
|
339
|
+
## Exchange #6
|
|
340
|
+
| From | To | Type | Timestamp |
|
|
341
|
+
|------|----|------|-----------|
|
|
342
|
+
| `tech-lead` | `all` | DECISION | {time} |
|
|
343
|
+
|
|
344
|
+
**Status:** โ
CONSENSUS REACHED
|
|
345
|
+
**Summary:** [Final output approved โ merged deliverable description]
|
|
346
|
+
**Signatures:** TechLead โ | Executor โ | Reviewer โ
|
|
347
|
+
```
|
|
348
|
+
|
|
349
|
+
**Mailbox types**: `TASK_ASSIGNMENT`, `SUBMISSION`, `REVIEW`, `DEFENSE`, `ESCALATION`, `APPROVAL`, `DECISION`
|
|
350
|
+
|
|
351
|
+
**Rules**:
|
|
352
|
+
- Append-only โ no agent may edit prior exchanges
|
|
353
|
+
- Every exchange MUST have From, To, Type, and Timestamp
|
|
354
|
+
- REVIEW exchanges MUST include explicit PASS or FAIL status
|
|
355
|
+
- DEFENSE exchanges MUST include technical evidence (not just disagreement)
|
|
356
|
+
- DECISION exchanges MUST include the consensus stamp
|
|
357
|
+
|
|
358
|
+
---
|
|
359
|
+
|
|
360
|
+
## TEAM PHASE OUTPUT FORMAT
|
|
361
|
+
|
|
362
|
+
```markdown
|
|
363
|
+
## ๐ญ Phase {N}: {name} โ ๐บ GOLDEN TRIANGLE
|
|
364
|
+
|
|
365
|
+
### Triangle: `{team-domain}`
|
|
366
|
+
**Tech Lead**: `{agent}` | **Executor**: `{agent}` | **Reviewer**: `{agent}`
|
|
367
|
+
|
|
368
|
+
### ๐ Task List
|
|
369
|
+
| ID | Task | Status | Round |
|
|
370
|
+
|----|------|--------|-------|
|
|
371
|
+
| T1 | {task} | โ
| 2 |
|
|
372
|
+
| T2 | {task} | โ
| 1 |
|
|
373
|
+
|
|
374
|
+
### ๐จ Executor Output
|
|
375
|
+
{Implementation summary โ what was built, approach taken, key decisions}
|
|
376
|
+
|
|
377
|
+
### ๐ Review Cycle
|
|
378
|
+
**Round 1**: โ FAIL โ {findings summary: what failed, severity}
|
|
379
|
+
**Round 2**: โ
PASS โ {resolution summary: what was fixed, what was defended}
|
|
380
|
+
|
|
381
|
+
### ๐ฌ Debate Log
|
|
382
|
+
{Key Mailbox exchanges โ defenses raised, escalations, resolutions. Brief, not full Mailbox.}
|
|
383
|
+
|
|
384
|
+
### ๐ค Consensus
|
|
385
|
+
โ
CONSENSUS: TechLead โ | Executor โ | Reviewer โ
|
|
386
|
+
**Decision**: {Final approved output description}
|
|
387
|
+
|
|
388
|
+
### Exit Criteria
|
|
389
|
+
- [x] {criterion_1} โ verified by `{agent}`
|
|
390
|
+
- [x] {criterion_2} โ verified by `{agent}`
|
|
391
|
+
|
|
392
|
+
### โ
Phase complete โ Golden Triangle approved
|
|
393
|
+
**Deliverable**: {consolidated summary of approved output}
|
|
394
|
+
```
|
|
395
|
+
|
|
396
|
+
---
|
|
397
|
+
|
|
398
|
+
## WHEN TO USE TEAMS
|
|
399
|
+
|
|
400
|
+
### Decision Tree
|
|
401
|
+
|
|
402
|
+
```
|
|
403
|
+
IS task complex with multiple concerns (security + performance + correctness)?
|
|
404
|
+
โโ YES โ :team (Golden Triangle)
|
|
405
|
+
โโ NO
|
|
406
|
+
IS quality critical and adversarial review needed?
|
|
407
|
+
โโ YES โ :team (Golden Triangle)
|
|
408
|
+
โโ NO
|
|
409
|
+
IS task simple and single-domain?
|
|
410
|
+
โโ YES โ Single-agent variant (:fast, :hard, :focus)
|
|
411
|
+
โโ NO โ Single-agent variant with manual review step
|
|
412
|
+
```
|
|
413
|
+
|
|
414
|
+
### Variant Comparison
|
|
415
|
+
|
|
416
|
+
| Variant | Execution Mode | Agents | When |
|
|
417
|
+
|---------|----------------|--------|------|
|
|
418
|
+
| `:fast` | Single agent | 1 | Speed priority, simple tasks |
|
|
419
|
+
| `:hard` | Single agent | 1 | Standard quality, focused tasks |
|
|
420
|
+
| `:focus` | Single agent + context gate | 1 | Clean execution, noise reduction |
|
|
421
|
+
| `:team` | Golden Triangle | 3 | Maximum quality, adversarial review, complex tasks |
|
|
422
|
+
|
|
423
|
+
**โ Do NOT use `:team` for**:
|
|
424
|
+
- Simple single-domain tasks where one agent suffices
|
|
425
|
+
- Time-critical hotfixes where coordination overhead exceeds benefit
|
|
426
|
+
- Pure research or brainstorming with no reviewable deliverable
|
|
427
|
+
|
|
428
|
+
---
|
|
429
|
+
|
|
430
|
+
## TEAM SIZE
|
|
431
|
+
|
|
432
|
+
**Always exactly 3.** No more, no less.
|
|
433
|
+
|
|
434
|
+
| Constraint | Value | Rationale |
|
|
435
|
+
|------------|-------|-----------|
|
|
436
|
+
| Tech Lead count | 1 | Single point of authority and synthesis |
|
|
437
|
+
| Executor count | 1 | Single implementer (except `fullstack` = 2 taking turns) |
|
|
438
|
+
| Reviewer count | 1 | Single quality gatekeeper with focused expertise |
|
|
439
|
+
| Total agents | 3 | Minimum viable adversarial collaboration |
|
|
440
|
+
|
|
441
|
+
```
|
|
442
|
+
IF only 1 perspective needed โ Single agent, not a team
|
|
443
|
+
IF 4+ perspectives needed โ Run multiple Golden Triangles sequentially per phase
|
|
444
|
+
(each triangle handles one concern domain)
|
|
445
|
+
NEVER spawn 4+ agents in one triangle
|
|
446
|
+
```
|
|
447
|
+
|
|
448
|
+
---
|
|
449
|
+
|
|
450
|
+
## ERROR HANDLING
|
|
451
|
+
|
|
452
|
+
| Error | Symptom | Recovery |
|
|
453
|
+
|-------|---------|----------|
|
|
454
|
+
| Executor fails to produce deliverable | SUBMISSION never posted to Mailbox | Tech Lead posts DECISION: phase blocked. Re-dispatches task or produces minimal viable output. |
|
|
455
|
+
| Reviewer is too strict (fails valid work 3 times) | Max rounds hit, all submissions rejected | Tech Lead reads all evidence, makes binding DECISION. Documents why Reviewer's standards were overridden. |
|
|
456
|
+
| Reviewer rubber-stamps (passes without scrutiny) | REVIEW contains no specific findings | Tech Lead rejects the PASS, re-invokes Reviewer with explicit checklist: security, performance, correctness, edge cases. |
|
|
457
|
+
| Consensus impossible after max rounds | 3 rounds exhausted, still FAIL | Tech Lead makes binding DECISION based on cumulative Mailbox evidence. Logs "ARBITRATED โ no consensus." |
|
|
458
|
+
| TIER 1 spawn fails for any agent | Sub-agent creation error | Fall back to TIER 2 (sequential embodiment) for the failed agent only. Log reason. |
|
|
459
|
+
| Mailbox file cannot be created | File system error | Use inline communication within the phase output. Log degraded mode. |
|
|
460
|
+
| All agents fail | No usable output from any role | Tech Lead produces minimal viable output solo, flags incident for Orchestrator review. |
|
|
461
|
+
|
|
462
|
+
---
|
|
463
|
+
|
|
464
|
+
## INTEGRATION WITH EXISTING RULES
|
|
465
|
+
|
|
466
|
+
| Rule File | Integration Point |
|
|
467
|
+
|-----------|-------------------|
|
|
468
|
+
| `CORE.md` | Golden Triangle respects all 10 Orchestration Laws. Tech Lead is bound by L7 (meta agents delegate). Executor and Reviewer follow L3 (complete assigned scope). |
|
|
469
|
+
| `PHASES.md` | Golden Triangle output extends phase format. Exit criteria verification unchanged. Each phase can use a triangle. |
|
|
470
|
+
| `AGENTS.md` | All three roles follow agent categories and context isolation rules. TIER 1/2 applies per-agent within the triangle. |
|
|
471
|
+
| `SKILLS.md` | Executor resolves skills independently via HSOL for implementation. Reviewer resolves skills for review checklists. Tech Lead does NOT resolve skills โ only coordinates. |
|
|
472
|
+
| `ERRORS.md` | Triangle errors follow standard error recovery. Escalation path: Executor โ Reviewer โ Tech Lead โ Orchestrator. |
|
|
473
|
+
|
|
474
|
+
---
|
|
475
|
+
|
|
476
|
+
## SELF-CHECK (Before Every Team Phase)
|
|
477
|
+
|
|
478
|
+
```
|
|
479
|
+
โก Is :team variant explicitly invoked? (Never use Golden Triangle without request)
|
|
480
|
+
โก Is the correct team domain identified from the roster?
|
|
481
|
+
โก Are exactly 3 roles assigned (Tech Lead + Executor + Reviewer)?
|
|
482
|
+
โก Has Tech Lead produced the Shared Task List BEFORE dispatch?
|
|
483
|
+
โก Is the Mailbox file initialized at ./reports/MAILBOX-{date}.md?
|
|
484
|
+
โก Is TIER 1 attempted first? (TIER 2 only on spawn failure)
|
|
485
|
+
โก Is the debate loop capped at 3 rounds?
|
|
486
|
+
โก Does every REVIEW contain explicit PASS/FAIL?
|
|
487
|
+
โก Does every DEFENSE contain technical evidence?
|
|
488
|
+
โก Is the consensus stamp present before phase output is released?
|
|
489
|
+
โก Does the final deliverable meet all exit criteria?
|
|
490
|
+
โก Are all Mailbox exchanges preserved (append-only, no deletions)?
|
|
491
|
+
```
|
|
492
|
+
|
|
493
|
+
---
|
|
494
|
+
|
|
495
|
+
## GOLDEN TRIANGLE vs. OLD TEAM MODEL
|
|
496
|
+
|
|
497
|
+
| Aspect | Old Model (Team Lead + Teammates) | Golden Triangle |
|
|
498
|
+
|--------|-----------------------------------|-----------------|
|
|
499
|
+
| Team size | 3-6 agents | Always exactly 3 |
|
|
500
|
+
| Quality mechanism | Team Lead review only | Adversarial Executor vs. Reviewer debate |
|
|
501
|
+
| Conflict resolution | Team Lead decides, no debate | Structured 3-round debate with evidence |
|
|
502
|
+
| Roles | Flexible teammates, no fixed roles | Fixed: Tech Lead, Executor, Reviewer |
|
|
503
|
+
| Communication | Ad-hoc Mailbox messages | Structured exchange types with PASS/FAIL |
|
|
504
|
+
| Consensus | Implicit (Team Lead merges) | Explicit stamp required from all 3 agents |
|
|
505
|
+
| Failure mode | Silent quality gaps | Reviewer catches gaps OR Tech Lead arbitrates |
|
|
506
|
+
| Defense mechanism | None โ teammates accept all feedback | Executor MUST defend valid work with evidence |
|
|
507
|
+
|
|
508
|
+
---
|
|
509
|
+
|
|
510
|
+
## QUICK REFERENCE
|
|
511
|
+
|
|
512
|
+
```
|
|
513
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
514
|
+
โ GOLDEN TRIANGLE โ CHEAT SHEET โ
|
|
515
|
+
โ โ
|
|
516
|
+
โ ALWAYS 3 AGENTS: Tech Lead ยท Executor ยท Reviewer โ
|
|
517
|
+
โ MAX 3 DEBATE ROUNDS: Implement โ Review โ Fix/Defend โ
|
|
518
|
+
โ MAILBOX: ./reports/MAILBOX-{date}.md (append-only) โ
|
|
519
|
+
โ CONSENSUS: โ
TechLead โ | Executor โ | Reviewer โ โ
|
|
520
|
+
โ โ
|
|
521
|
+
โ Tech Lead = FINAL AUTHORITY (arbitrates, synthesizes) โ
|
|
522
|
+
โ Executor = BUILDER + DEFENDER (implements, pushes back) โ
|
|
523
|
+
โ Reviewer = DEVIL'S ADVOCATE (challenges, validates) โ
|
|
524
|
+
โ โ
|
|
525
|
+
โ NO output without consensus stamp โ
|
|
526
|
+
โ NO defense without technical evidence โ
|
|
527
|
+
โ NO rubber-stamp reviews (Tech Lead rejects empty passes) โ
|
|
528
|
+
โ NO more than 3 rounds (Tech Lead decides at round 3) โ
|
|
529
|
+
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
|
|
530
|
+
```
|