codeconscious-identity
Explore codebases with persistent memory
Understanding large codebases requires time and effort. CodeConscious provides an autonomous cognitive assistant that builds persistent memory across sessions, enabling deep code exploration, pattern recognition, and continuous learning from project history.
下載技能 ZIP
在 Claude 中上傳
前往 設定 → 功能 → 技能 → 上傳技能
開啟並開始使用
測試它
正在使用「codeconscious-identity」。 /runtime.explore
預期結果:
- Technology stack: Node.js + Express + PostgreSQL + Redis
- Architecture pattern: Layered architecture (API → Service → Repository)
- Key modules: auth/service.js (centrality 0.85), user/repository.js (0.78)
- Dependencies: 47 files, 132 relationships identified
- Recommendations: Add tests for auth/service.js, refactor utils/helpers.js
正在使用「codeconscious-identity」。 /runtime.think "Why is the payment service slow"
預期結果:
- Analysis of payment service latency: 3 hypotheses identified
- Hypothesis 1: Database connection pool exhaustion (confidence 0.75)
- Hypothesis 2: Missing Redis cache layer for frequently accessed data (confidence 0.70)
- Hypothesis 3: Synchronous logging blocking async operations (confidence 0.60)
- Recommended: Check connection pool configuration first
正在使用「codeconscious-identity」。 /runtime.learn "Docker container security best practices"
預期結果:
- Learning complete with confidence 0.92
- Key patterns identified: minimal base images, non-root users, read-only filesystems
- Memory updated: security patterns for container deployment
- Sources reviewed: 8 documentation pages, 3 security guides
安全審計
安全Pure documentation skill containing only markdown files defining AI command behaviors. All 651 static findings are FALSE POSITIVES because: (1) files are markdown documentation, not executable code; (2) bash command examples in docs are not vulnerabilities; (3) SHA-256 hash fields are secure, not weak algorithms; (4) path examples in docs are not path traversal; (5) documentation links are legitimate. No actual code execution, network calls, filesystem access, or credential handling exists.
風險因素
⚙️ 外部命令 (553)
品質評分
你能建構什麼
Understand large codebases
Explore new projects quickly by building cognitive maps and dependency graphs automatically
Maintain project knowledge
Preserve architectural decisions and patterns in persistent memory across team members
Build cognitive AI agents
Implement autonomous learning agents with confidence tracking and memory systems
試試這些提示
/runtime.explore - scan this codebase, identify technology stack, architecture patterns, and key modules
/runtime.learn "why does the payment service timeout under load" - explore code, check logs, find root cause
/runtime.plan "implement OAuth2.0 authentication with JWT tokens" - break down into tasks with dependencies
/runtime.remember "we decided to use PostgreSQL for all new projects because of its JSON support"
最佳實務
- Use /runtime.explore first when working with unfamiliar codebases to build context
- Always use /runtime.learn before planning to understand existing architecture
- Mark confidence levels explicitly for all recommendations and findings
- Record architectural decisions with /runtime.remember for future team reference
避免
- Starting implementation without exploring or learning the codebase first
- Skipping confidence annotations on architectural recommendations
- Forgetting to record important decisions that affect future development
- Making changes without checking existing patterns in memory