ECC Batch-3 — Runtime Hooks Sandbox Audit
Repository evidence: research/agent-architecture/audits/2026-09-27-ecc-batch-3-hooks-sandbox-audit.md
Atlas Bidang
Arsitektur Agen
Agen terbatas, alat, konteks, memori, keandalan, evaluasi, suara, pengambilan informasi, dan perluasan fungsi.
45 kapabilitas dipetakan ke kelompok tematik agar pembaca dapat bergerak dari ruang masalah menuju pilihan implementasi.
Peta kapabilitas
Peta capability Arsitektur Agen
Penalaran dan pengarahan model
Pemilihan model, pilihan cadangan, dan batas penalaran.
Konteks dan memori
Rekayasa konteks, memori persisten, dan keadaan tugas.
Keandalan, evaluasi, dan keamanan
Recovery, evals, authority, approval, dan safe operation.
Pengambilan informasi dan pengetahuan
RAG, retrieval, embeddings, dan external knowledge integration.
Alat, suara, dan multimodal
Penggunaan alat, perluasan fungsi, suara, pengamatan multimodal, dan interaksi komputer.
Kapabilitas lainnya
Catatan bidang yang belum masuk kelompok tematik utama.
- Bounded Agent Harness Policy
- Conditional Multi-Agent Coordination Policy
- ECC GateGuard Fact-Forcing Gate Hook
- ECC Git Hook Bypass Guard Hook
- ECC Linter Config Protection Hook
- ECC Market Research Evidence Standards
- ECC Plugin Wholesale Runtime Install (Rejected)
- ECC Prediction Market Risk Review Reference
- Google Gemma Development Skill
- Mark-LII Personal AI Assistant Architecture Reference
- OpenTelemetry Specification
- Proactive and Event-Driven Agent Execution Policy
- Temporal
Bukti riset
Bukti yang mendukung bidang ini
Claude Code Five-Capability Runtime Differential Audit
Repository evidence: research/agent-architecture/audits/2026-09-03-claude-code-five-capability-runtime-audit.md
Context7 Candidate Audit
Repository evidence: research/agent-architecture/audits/2026-08-28-context7-candidate-audit.md
Agent Architecture Deep Audit — Batch A: Harness and Tool Routing
Repository evidence: research/agent-architecture/audits/2026-08-26-batch-a-harness-tool-routing.md
Agent Architecture Deep Audit — Batch B: Context Engineering and Memory
Repository evidence: research/agent-architecture/audits/2026-08-26-batch-b-context-memory.md
Agent Architecture Deep Audit — Batch C: Planning, State, Resumability and Multi-Agent
Repository evidence: research/agent-architecture/audits/2026-08-26-batch-c-planning-state-multi-agent.md
Agent Architecture Deep Audit — Batch D: Security, Authority, Reliability and Recovery
Repository evidence: research/agent-architecture/audits/2026-08-26-batch-d-security-authority-reliability.md
Agent Architecture Deep Audit — Batch E: Evaluation and Observability
Repository evidence: research/agent-architecture/audits/2026-08-26-batch-e-evals-observability.md