c24165b8bc
에이전트 8종(대화형/프로그래머 C·S/디자이너/설계자/기획자/QA/PD/PM)에게 [공통 기본 능력 + 롤별 Specialty] 2층으로 지식을 주입하기 위한 재분류. 문서 내용·포맷은 무수정, 폴더 이동만 (6,372개 문서 수 보존 확인). - Topic_Programming → Domain_Programming (내부 구조 보존) - Topic_Graphic → Domain_Design - Topic_Business → Domain_Product - Topic_General → Domain_General - _Common 신설: Math(구 Topic_Math_Specialty), Reasoning(구 General/From_Thinking & Reasoning), Reasoning_Creativity(구 General/From_창의성), Communication(Poetic_Blog_Writing + From_writing) - 타 도메인의 From_* 폴더는 유지 (출처 표기일 뿐, 이미 도메인에 맞게 분류된 문서) - 빈 폴더 정리 (memory/procedures) - 에이전트→폴더 매핑은 workspace의 .astra/agent-knowledge-map.json (9개 에이전트) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
4.3 KiB
4.3 KiB
id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack
| id | title | category | status | canonical_id | aliases | duplicate_of | source_trust_level | confidence_score | verification_status | tags | raw_sources | last_reinforced | github_commit | tech_stack | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| wiki-2026-0508-restorative-justice | Restorative Justice | 10_Wiki/Topics | verified | self |
|
none | A | 0.85 | applied |
|
2026-05-10 | pending |
|
Restorative Justice
매 한 줄
"매 처벌이 아닌 관계 회복으로 갈등을 해결". 매 1970s Howard Zehr 의 정립, 매 New Zealand 1989 Family Group Conference 법제화 가 milestone. 매 2026 software/community governance, AI safety incident response 에도 적용.
매 핵심
매 vs Punitive Justice
- Punitive: "What law was broken? Who did it? What punishment?"
- Restorative: "Who was harmed? What are their needs? Whose obligations are these?"
- Procedural: rule-following 강조. RJ 는 outcome (harm repair) 강조.
매 핵심 가치
- Accountability: 책임 의 active 인정 — passive 처벌 X.
- Healing: victim, offender, community 모두 의 회복.
- Voice: stakeholder 모두 의 발언권.
- Relationship: 관계 회복 우선.
매 응용
- School discipline (US, UK 의 도입).
- Criminal justice diversion programs.
- Workplace conflict (HR mediation).
- Online community moderation (Discord, GitHub).
- AI incident response (post-mortem culture).
💻 패턴
Conference structure
1. Pre-conference: 매 stakeholder 와 facilitator 의 prep.
2. Storytelling: 매 each party 의 narrative.
3. Impact: harm 의 articulation.
4. Need identification: what would help?
5. Agreement: concrete actions.
6. Follow-up: accountability check.
Software incident (post-mortem)
# Post-Mortem: Production Outage 2026-05-09
## Who was affected
- 12,000 users (3hr API down)
- On-call engineer (sleep disruption)
- CS team (200 tickets)
## What happened (no blame)
- Deploy at 03:00 UTC introduced N+1 query
- Monitoring alert delayed 15min
## Repair
- [ ] Refund/credit affected users
- [ ] On-call rotation adjustment
- [ ] CS team additional comp
## Prevention (system, not person)
- [ ] Pre-deploy load test gate
- [ ] Alert threshold tightening
Community moderation (RJ flow)
type Incident = {
reporter: UserId;
accused: UserId;
description: string;
};
async function restorativeFlow(i: Incident) {
await dm(i.reporter, "We hear you. What outcome would help?");
const need = await waitReply(i.reporter);
await dm(i.accused, `A community member shared they were hurt by [...]. Are you open to a conversation?`);
if (await consent(i.accused)) {
await scheduleFacilitatedChat(i.reporter, i.accused, need);
} else {
await escalateToFormal(i);
}
}
Circle process
- 매 participant 의 talking piece 보유시 발언.
- 매 facilitator 의 question prompt 의 sequence.
- 매 silence OK.
- 매 agreement 의 collective.
Apology framework (Lazare)
1. Acknowledgment: "I did X."
2. Explanation (not excuse): "Because Y."
3. Remorse: "I see harm caused."
4. Repair: "I will Z to make it right."
매 결정 기준
| 상황 | Approach |
|---|---|
| First-time, low-harm | RJ conference |
| Repeated, predatory | Punitive + protection |
| Power imbalance severe | RJ X (re-traumatization risk) |
| Anonymous online harm | Hybrid (account action + community education) |
| Workplace senior→junior | RJ + structural change |
기본값: harm-first framing + voluntary participation + structural prevention.
🔗 Graph
🤖 LLM 활용
언제: post-mortem template 작성, mediation script 초안, policy review. 언제 X: live mediation (human empathy 의 irreplaceable).
❌ 안티패턴
- Forced participation: 매 victim 의 secondary harm.
- Surface-only: structural cause 의 외면.
- No follow-up: agreement 후 accountability gap.
- Power blind: senior↔junior 의 same-level treatment.
🧪 검증 / 중복
- Verified (Howard Zehr "Changing Lenses", UN handbook 2020).
- 신뢰도 A.
🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — RJ + software application |