القائمة
Agent reputation scoring, performance tier system, trust calibration, and task affinity matching
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
استنادا إلى تصنيف SOC المهني
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/vibeeval/vibecosystem --skill reputation-patternsيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
عرض SKILL.md
Claude Code CLI commands, flags, headless mode, and automation patterns
OpenAI Codex CLI + Claude Code (Hizir) birlikte kullanim rehberi. Is dagitim pattern'leri, GitHub Actions workflow ornekleri, review dongusu ve iki AI yazilim asistaninin guclu yanlarini birlestiren orchestration stratejileri.
Meta-skill for internal codebase exploration at varying depths (quick/deep/architecture)
| name | reputation-patterns |
| description | Agent reputation scoring, performance tier system, trust calibration, and task affinity matching |
interface AgentReputation {
agentId: string
score: number // 0-100
tier: 'S' | 'A' | 'B' | 'C' | 'D'
totalTasks: number
successRate: number // 0-1
firstPassRate: number // QA'den ilk seferde geçme oranı
avgRetries: number
specialties: string[] // Başarılı olduğu task tipleri
weaknesses: string[] // Başarısız olduğu task tipleri
}
// Score hesaplama
function calculateARS(agent: AgentReputation): number {
const weights = {
successRate: 0.35,
firstPassRate: 0.25,
consistency: 0.20, // son 10 task varyansı
efficiency: 0.20 // ortalama retry sayısı
}
return (
agent.successRate * weights.successRate * 100 +
agent.firstPassRate * weights.firstPassRate * 100 +
(1 - agent.avgRetries / 3) * weights.consistency * 100 +
(agent.totalTasks > 5 ? 1 : 0.5) * weights.efficiency * 100
)
}
| Tier | ARS Range | Yetki | Açıklama |
|---|---|---|---|
| S | 90-100 | Otonom | Her task'ı alabilir, QA skip |
| A | 75-89 | Güvenilir | Çoğu task, standart QA |
| B | 60-74 | Normal | Standart task, tam QA |
| C | 40-59 | İzlemede | Basit task, sıkı QA |
| D | 0-39 | Kısıtlı | Cross-training gerekli |
Decay:
- 7 gün inaktif → -2 puan
- 14 gün inaktif → -5 puan
- QA FAIL → -5 puan (severity'ye göre)
- Escalation → -10 puan
Recovery:
- QA PASS (first try) → +3 puan
- QA PASS (retry) → +1 puan
- Complex task başarı → +5 puan
- Cross-training başarı → +2 puan
## Agent Affinity Matrix
| Agent | Frontend | Backend | DB | Security | Score |
|-------|----------|---------|----|---------|----|
| spark | 85 | 70 | 50 | 40 | B |
| kraken | 60 | 90 | 80 | 65 | A |
| frontend-dev | 95 | 30 | 20 | 30 | S(FE) |
| backend-dev | 20 | 92 | 75 | 60 | S(BE) |
Task assignment: Highest affinity agent with available capacity
def predict_success(agent_id: str, task_type: str) -> float:
"""Agent'ın belirli task tipinde başarı olasılığı"""
agent = get_agent(agent_id)
base_rate = agent.success_rate
type_affinity = agent.affinity.get(task_type, 0.5)
recent_trend = agent.last_5_tasks_success_rate
return (base_rate * 0.4 + type_affinity * 0.35 + recent_trend * 0.25)
# Skill matrix'ten data çek
cat ~/.claude/canavar/skill-matrix.json | jq '.agents[] | {name, score, tier}'
# Error ledger'dan failure pattern
cat ~/.claude/canavar/error-ledger.jsonl | jq 'select(.agent == "spark")'
# Leaderboard
node ~/.claude/hooks/dist/canavar-cli.mjs leaderboard