Novera-AI-skills

A curated collection of eight production-style Novera-AI-skills / Claude-style skills, selected from globally trending agent-skill categories: memory persistence, code quality, data analysis, debugging, testing, infrastructure-as-code security, performance, and documentation.

Every skill ships with runnable, dependency-free scripts (Python standard library only) so they work in CI and local shells without extra installs.

Skills

1. persistent-memory

File-based cross-session memory: remember facts/preferences, recall by topic, and periodically distill daily logs into a curated MEMORY.md.

Triggers: "记住", "记下来", "长期记忆", "别忘记", "recall", "remember", "memory", "跨会话".

2. code-review-security

Pre-merge code review and security audit using a five-axis framework (correctness, readability, architecture, security, performance) with severity classification and a standardized report template.

Triggers: "review 代码", "代码审查", "安全审计", "审查这个 PR", "review this PR", "security audit", "code review".

3. nl-data-analysis

Turn natural-language questions about tabular data into pandas/SQL analysis, charts, and plain-language insights.

Triggers: "分析这个表", "画个图", "这数据说明什么", "analyze this data", "make a chart", "数据可视化", "跑个分析".

4. agentic-debugger

Structured debugging loop for agent/subagent failures: parse a traceback, generate a minimal pytest reproduction, and run a pass/fail test guard that proves the fix. Turns "it broke" into a reproducible, verifiable cycle.

Triggers: "debug 这个报错", "复现这个 bug", "解析堆栈", "生成复现", "test guard", "debug this traceback", "reproduce the failure".

5. testforge

Generate pytest test stubs from source via AST analysis and surface coverage gaps with actionable hints. Speeds up test-authoring so coverage is real, not padded.

Triggers: "生成测试", "写单测", "pytest 桩", "覆盖率提示", "gen tests", "generate test stubs", "coverage hint".

6. iac-guard

Heuristic security scanner for Terraform / Kubernetes / YAML IaC: flags privileged containers, wildcard IAM, open security groups, missing resource limits, and more. CI-friendly exit codes and JSON output.

Triggers: "扫描 IaC", "检查 Terraform", "K8s 安全", "基础设施安全", "scan iac", "terraform security", "k8s guard".

7. perf-doctor

Static performance doctor for application code: detects N+1 queries, SELECT *, nested loops, missing LIMIT, and synchronous calls inside loops. Zero-config, CI-friendly, no runtime overhead.

Triggers: "性能检查", "慢查询", "N+1", "扫描性能问题", "perf check", "scan performance", "find N+1".

8. docsmith

Auto-draft docs from code: generate a README from module docstrings/structure and a CHANGELOG grouped by Conventional Commits from git history. Keeps docs in sync without manual busywork.

Triggers: "生成 README", "写文档", "生成 CHANGELOG", "从代码出文档", "gen readme", "generate changelog", "docs from code".

Structure

Each skill is a directory containing a SKILL.md (metadata + instructions) and optional scripts/, references/, assets/. All scripts/*.py use only the Python standard library — no pip install required.

Install

Copy any skill directory into your agent skills folder (e.g. ~/.novera-ai-skills/skills/ or your platform's equivalent), then invoke it by its trigger phrases.

License

MIT