# Deep Research

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-junfengran-dev-env-deep-research`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [JunfengRan](https://agentstack.voostack.com/s/junfengran)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [JunfengRan](https://github.com/JunfengRan)
- **Source:** https://github.com/JunfengRan/dev-env/tree/main/skills/deep-research
- **Website:** https://github.com/JunfengRan/dev-env

## Install

```sh
agentstack add skill-junfengran-dev-env-deep-research
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Deep Research（薄编排层）

**阶段定义不在本 SKILL** — 唯一真相源是 [`spec/research-workflow.yaml`](../../spec/research-workflow.yaml)。

## 启发式补充（非 Spec 阶段）

研究问题或选型边界仍模糊时，**可先**用 [`brainstorm`](../brainstorm/SKILL.md) 做几轮澄清与方案对比，再把结论写进 `research-brief.json`。

- brainstorm **不是** Spec state，不经 gate / reducer
- 用户已明确读者、scope、evidenceTargets 时可跳过
- 交接模板见 `brainstorm/references/handoff-templates.md`

## 启动 / 恢复 Run

1. 用 CLI 新建 run（推荐）或恢复已有目录（见 [`spec/run-directory.md`](../../spec/run-directory.md)）：

```bash
node scripts/research-cli.mjs init  --slug 
# 或: npm run research -- init  --slug 
```

2. 确认 `run-meta.json`、`state.json`、`context-pack.json`、`observations.jsonl`、`replay-chain.json` 已生成
3. Read Spec + 当前 `state.json` 的 `currentState`（也可用 `node scripts/research-cli.mjs status`）

## 每轮循环

```
读 Spec 当前 state → 按 contextSlice 组装 ContextPack 注入
→ 执行（agent / skill / parallel_subagents）
→ 写 artifacts → research-cli advance（gate + reducer + bump）
→ 写 observation（subagent 用 after-subagent-complete）
```

## 状态类型

| kind | 行为 |
|------|------|
| `agent` | 主 deep-research agent 执行 |
| `parallel_subagents` | dispatch codebase-verifier；barrier 同步 |
| `skill` | 调用 interpret-tech-notes 或 write-idea-docs |
| `terminal` | 结束 |

## 禁止

- 未跑 gate / 未 `advance` 就宣称 phase 完成
- 跳过 Spec 未定义的 transition
- 子 agent 结论只写 chat 不写 `observations.jsonl`
- 手改 `state.json` barrier（应走 reducer / CLI / after-subagent-complete）

## References

- [read-spec-guide.md](references/read-spec-guide.md)
- [context-pack-guide.md](references/context-pack-guide.md)
- [subagent-dispatch-guide.md](references/subagent-dispatch-guide.md)
- [replay-guide.md](references/replay-guide.md)
- [plan-critique-playbook.md](references/plan-critique-playbook.md)
- [depth-checklist-decision.md](references/depth-checklist-decision.md)

## 工具命令

```bash
node scripts/research-cli.mjs init  --slug 
node scripts/research-cli.mjs status [run-dir]
node scripts/research-cli.mjs advance [run-dir]
node scripts/research-cli.mjs apply  ''
node scripts/gates/.mjs .research/
node plugins/deep-research-gates/scripts/after-subagent-complete.mjs .research/  
node plugins/deep-research-gates/scripts/bump-context-pack.mjs .research/ 
npm run validate:research-spec
```

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [JunfengRan](https://github.com/JunfengRan)
- **Source:** [JunfengRan/dev-env](https://github.com/JunfengRan/dev-env)
- **License:** MIT
- **Homepage:** https://github.com/JunfengRan/dev-env

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-junfengran-dev-env-deep-research
- Seller: https://agentstack.voostack.com/s/junfengran
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.