shieldcortex-skill
jarvis-drakon 제작
Notes & PKM요약
Give your AI agent a brain that persists between sessions — and protect it from memory poisoning attacks.
설명
이 스킬이 하는 일
ShieldCortex is a persistent memory system for AI agents with a built-in 6-layer security pipeline. It stores memories in SQLite with semantic search, knowledge graph extraction, decay-based forgetting, and contradiction detection. Every memory write is screened for prompt injection, credential leaks, and poisoning attacks before being saved.
Unlike raw context windows or simple key-value stores, ShieldCortex combines semantic retrieval with active security screening, so agents can recall relevant past context without becoming a target for memory poisoning.
사용 시점
- Remembering architectural decisions across coding sessions
- Detecting prompt injection in installed skill files before running them
- Auditing what an agent has stored and retrieved over time
- Blocking accidental credential writes to agent memory
- Building a knowledge graph of entities from project notes
예시 워크플로
User asks: remember this: the payment API requires OAuth2 bearer tokens, not API keys
- 1
Trigger fires on 'remember this:' keyword in session
- 2
Content passes through 6-layer defence pipeline checking for injection patterns and credential leaks
- 3
Memory is saved to SQLite with category 'architecture' and importance level
- 4
Knowledge graph extraction identifies entities: payment API, OAuth2, API keys
- 5
Memory becomes retrievable via semantic search in future sessions
Memory stored and confirmed; entities added to knowledge graph
요구 사항
- ShieldCortex Cloud API key (optional, for team dashboards and audit aggregation)
설치
clawhub install jarvis-drakon/shieldcortex-skillCLI 명령을 실행하거나, 프롬프트를 Claude Code, Cursor, Codex 등 사용하는 코딩 에이전트에 붙여넣으세요.