据 AlphaMoat 统计,截至 2026-10,ResearchVault 累计下载 5,612 次、获 6 个星标,按累计下载量在本次参与排名的 10,914 个 AI 技能中排名第 5,140 位,在「AI 智能体」分类(1,632 个)中排名第 714 位。
列表默认按累计下载量排序,也可按星标排序;详情页排名按累计下载量计算。
| 平台 | 累计下载 | 星标 | 评论 |
|---|---|---|---|
| ClawHub ↗ | 5,584 | 6 | - |
| SkillHub ↗ | 5,612 | 6 | — |
各平台的计数口径不同,统计时间不一致。
文档说明它管理持久状态、信息综合与自动验证,数据存储在本地 SQLite 数据库;搜索和后台服务可按需配置。
资料发布于 2026-10-27
以下是根据技能文档生成的示例,不代表实际运行结果。
2026-09 起统计方法调整:同名技能按作者分别统计,并覆盖全部技能。图中虚线为调整前的月份,与之后的数据不直接比较,也不计算跨越调整的增长。
| 月份 | 累计下载 | 环比 | 星标 |
|---|---|---|---|
| 2026-10 | 5,612 | — | 6 |
| 2026-09 | 5,411 | — | 6 |
| 2026-08 | 5,273 | +3.0% | 6 |
| 2026-07 | 5,117 | +4.9% | 6 |
| 2026-06 | 4,880 | +3.6% | 6 |
| 2026-05 | 4,709 | — | 6 |
以下是作者发布的原始文件,按纯文本显示,本站未做修改或审核;其中的链接不可点击,命令请核实后再运行。抓取于 2026-10-23。
---
name: researchvault
description: "Local-first research orchestration engine. Manages state, synthesis, and optional background services (MCP/Watchdog)."
homepage: https://github.com/lraivisto/ResearchVault
disable-model-invocation: true
user-invocable: true
metadata:
openclaw:
emoji: "🦞"
requires: python: ">=3.13"
env:
RESEARCHVAULT_DB:
description: "Optional: Custom path to the SQLite database file."
required: false
BRAVE_API_KEY:
description: "Optional: Brave Search API key."
required: false
SERPER_API_KEY:
description: "Optional: Serper API key."
required: false
SEARXNG_BASE_URL:
description: "Optional: SearXNG base URL."
required: false
RESEARCHVAULT_PORTAL_TOKEN:
description: "Optional: static portal token. If unset, start_portal.sh sources/generates .portal_auth and exports this env var."
required: false
RESEARCHVAULT_PORTAL_ALLOWED_DB_ROOTS:
description: "Optional: comma-separated absolute DB roots. Default: ~/.researchvault,/tmp."
required: false
RESEARCHVAULT_PORTAL_STATE_DIR:
description: "Optional: portal state directory (default ~/.researchvault/portal)."
required: false
RESEARCHVAULT_PORTAL_HOST:
description: "Optional: backend bind host."
required: false
RESEARCHVAULT_PORTAL_PORT:
description: "Optional: backend bind port."
required: false
RESEARCHVAULT_PORTAL_FRONTEND_HOST:
description: "Optional: frontend bind host."
required: false
RESEARCHVAULT_PORTAL_FRONTEND_PORT:
description: "Optional: frontend bind port."
required: false
RESEARCHVAULT_PORTAL_CORS_ORIGINS:
description: "Optional: comma-separated CORS origins for backend."
required: false
RESEARCHVAULT_PORTAL_RELOAD:
description: "Optional: set to 'true' for backend auto-reload."
required: false
RESEARCHVAULT_PORTAL_COOKIE_SECURE:
description: "Optional: set to 'true' to mark auth cookie Secure."
required: false
RESEARCHVAULT_PORTAL_PID_DIR:
description: "Optional: start_portal.sh PID/log directory."
required: false
RESEARCHVAULT_PORTAL_SHOW_TOKEN:
description: "Optional: set to '1' to print tokenized portal URLs."
required: false
RESEARCHVAULT_SEARCH_PROVIDERS:
description: "Optional: search provider order override."
required: false
RESEARCHVAULT_WATCHDOG_INGEST_TOP:
description: "Optional: watchdog ingest top-k override."
required: false
RESEARCHVAULT_VERIFY_INGEST_TOP:
description: "Optional: verify ingest top-k override."
required: false
RESEARCHVAULT_MCP_TRANSPORT:
description: "Optional: MCP server transport override."
required: false
REQUESTS_CA_BUNDLE:
description: "Optional: custom CA bundle for HTTPS verification."
required: false
SSL_CERT_FILE:
description: "Optional: custom CA certificate file."
required: false
---
# ResearchVault 🦞
**Local-first research orchestration engine.**
ResearchVault manages persistent state, synthesis, and autonomous verification for agents.
## Security & Privacy (Local First)
- **Local Storage**: All data is stored in a local SQLite database (~/.researchvault/research_vault.db). No cloud sync.
- **Network Transparency**: Outbound connections occur ONLY for user-requested research or Brave Search (if configured).
- **SSRF Hardening**: Strict internal network blocking by default. Local/private IPs (localhost, 10.0.0.0/8, etc.) are blocked. Use `--allow-private-networks` to override.
- **Manual Opt-in Services**: Background watchers and MCP servers are in `scripts/services/` and must be started manually.
- **Strict Control**: `disable-model-invocation: true` prevents the model from autonomously starting background tasks.
## Installation
```bash
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
```
## Quick Start
1. **Initialize Project**:
```bash
python scripts/vault.py init --objective "Analyze AI trends" --name "Trends-2026"
```
2. **Ingest Data**:
```bash
python scripts/vault.py scuttle "https://example.com" --id "trends-2026"
```
3. **Autonomous Strategist**:
```bash
python scripts/vault.py strategy --id "trends-2026"
```
## Portal (Manual Opt-In)
Start the portal explicitly:
```bash
./start_portal.sh
```
- Backend: `127.0.0.1:8000`
- Frontend: `127.0.0.1:5173`
- Backend auth strictly uses `RESEARCHVAULT_PORTAL_TOKEN`
- `./start_portal.sh` loads/generates `.portal_auth` and exports `RESEARCHVAULT_PORTAL_TOKEN` before backend launch
- Token login: URL hash `#token=<token>` (token from `.portal_auth`)
- Allowed DB roots are controlled by `RESEARCHVAULT_PORTAL_ALLOWED_DB_ROOTS` (default `~/.researchvault,/tmp`)
- OpenClaw workspace DBs are never discovered or selectable in Portal mode
- Provider secrets are env-only (`BRAVE_API_KEY`, `SERPER_API_KEY`, `SEARXNG_BASE_URL`) and are not injected into vault subprocesses
- Both hosts are supported for browser access:
- `http://127.0.0.1:5173/#token=<token>`
- `http://localhost:5173/#token=<token>`
Operational commands:
```bash
./start_portal.sh --status
./start_portal.sh --stop
```
Security parity with CLI:
- SSRF blocking is on by default (private/local/link-local targets denied).
- Portal toggle **Allow private networks** is equivalent to CLI `--allow-private-networks`.
## Optional Services (Manual Start)
- **MCP Server**: `python scripts/services/mcp_server.py`
- **Watchdog**: `python scripts/services/watchdog.py --once`
## Provenance & Maintenance
- **Maintainer**: lraivisto
- **License**: MIT
- **Issues**: [GitHub Issues](https://github.com/lraivisto/ResearchVault/issues)
- **Security**: See [SECURITY.md](SECURITY.md)
把开发过程中的经验、错误与纠正记录为 Markdown 文件,供后续复用并晋升为工作区记忆。
用自然语言描述需求,在本地与远程技能市场里找技能并给出推荐的技能。
给 AI agent 用的浏览器自动化命令行技能。
把用户纠正与自我复盘沉淀成分层记忆、并按使用情况升降级的技能。
一份为 AI 代理规划主动跟进、工作记忆与复盘流程的工作区配置指南。
指导如何创建与更新有效 Claude 技能的写作指南。
在安装第三方技能之前,按清单审查来源、权限与危险模式的技能。
把盛唐诗人李白的诗学观、意象系统与口吻,蒸馏成 AI 可调用的角色包。
数据月份:2026-10 · 下载/星标数聚合自公开技能注册库(ClawHub、SkillHub),按月更新。口径说明见数据方法。