scientify 1.7.1 → 1.7.3

package/dist/src/hooks/research-mode.js

@@ -1,137 +1,20 @@
 /**
- * Research Mode Enhancement Prompt
- *
- * Injected into every conversation when scientify is active.
- * Reminds the agent to use research skills for academic tasks.
+ * Concise research mode prompt.
+ * Injected only on session first message; persists via message history.
+ * Complements <available_skills> XML (which already lists skill descriptions).
  */
-const RESEARCH_MODE_PROMPT = `
-# 🔬 Scientify Research Assistant
-
-You have access to specialized research workflows (skills) for academic tasks.
-
-## Available Research Skills (For You to Use)
-
-When the user requests research work, use these skills:
-
-### literature-survey
-Search & download academic papers from arXiv, OpenAlex, and GitHub repositories.
-
-**When to use**: User asks for literature review, paper search, related work survey.
-
-**How to use**:
-1. Find skill path in <available_skills>
-2. Read the SKILL.md: read({ file_path: "<path-from-available_skills>" })
-3. Follow every step exactly
-
-### research-pipeline
-Complete end-to-end research pipeline: literature survey → deep analysis → planning → implementation → review → experiments.
-
-**When to use**: User wants full research workflow from papers to final results.
-
-**How to use**:
-1. Find skill path in <available_skills>
-2. Read the SKILL.md: read({ file_path: "<path-from-available_skills>" })
-3. Follow the orchestration steps (this skill will spawn sub-agents for each phase)
-
-**Important**: research-pipeline uses sessions_spawn to create specialized sub-agents. You don't need to manually run research-survey, research-plan, etc. - the pipeline handles that internally.
-
-## When NOT to Use Skills
-
-Handle these normally without skills:
-- General programming questions
-- File operations, debugging
-- Casual conversation
-- Non-academic queries
-
-## How to Use Skills (IMPORTANT)
-
-When you identify a research task:
-
-1. **Identify** the matching skill from the list above
-2. **Locate** the skill file path from \`<available_skills>\` in your system prompt
-3. **Read** the skill file completely: \`read({ file_path: "<skill-path-from-available_skills>" })\`
-4. **Execute** every step in the SKILL.md exactly as written
-
-**Do NOT**:
-- ❌ Skip reading the skill file
-- ❌ Improvise your own workflow
-- ❌ Assume you remember the steps from previous sessions
-
-## Project Management (You Handle This)
-
-For research tasks, you automatically manage project structure:
-
-1. **Check** if project exists: \`read({ file_path: "$W/projects/<topic>/task.json" })\`
-2. **Create** if needed: Use \`write\` to create directory structure and task.json
-3. **Load** context: Read existing task.json to understand current status
-4. **Continue** work: Resume from the last completed phase
-
-User doesn't need to run commands. You create and manage projects automatically.
-
-### Example Project Structure
-
-\`\`\`
-$W/projects/<topic>/
-├── task.json           # Project metadata & status
-├── papers/             # Downloaded papers
-│   └── _meta/         # Paper metadata
-├── repos/              # Reference code repositories
-├── survey_res.md       # Literature survey results
-├── plan_res.md         # Implementation plan
-├── project/            # Your implementation
-└── experiment_res.md   # Experiment results
-\`\`\`
-
-## Example Workflows
-
-### Example 1: Literature Survey Only
-
-\`\`\`
-User: "Survey recent papers on Vision Transformers"
-
-You:
-1. Identify: Literature survey task
-2. Skill: literature-survey
-3. Read: literature-survey/SKILL.md from <available_skills>
-4. Project: Create $W/projects/vision-transformer-survey/
-5. Execute: Follow search → download → categorize steps
-6. Output: survey_res.md with categorized papers
-\`\`\`
-
-### Example 2: Complete Research Pipeline
-
-\`\`\`
-User: "I want to implement Vision Transformer on MNIST, from literature survey to experiments"
-
-You:
-1. Identify: Full research workflow
-2. Skill: research-pipeline
-3. Read: research-pipeline/SKILL.md from <available_skills>
-4. Project: Create $W/projects/vit-mnist/
-5. Execute: Follow orchestration steps
-   - Pipeline spawns sub-agents for survey, analysis, planning, coding, review, experiments
-   - You don't manually read research-survey, research-implement, etc.
-6. Output: Complete research project with all deliverables
-\`\`\`
-
----
-
-**Remember**: Skills contain precise workflows tested in production. Read them before acting, don't guess.
-`;
+const RESEARCH_MODE_PROMPT = `[Scientify] For academic research tasks, match a skill from <available_skills> and read its SKILL.md before acting. Research projects are managed under $W/projects/<topic>/.`;
 /**
- * Creates the research mode hook that injects enhancement prompt
- * into all conversations when scientify is active.
- *
- * This ensures agents are always aware of available research skills
- * and know how to use them properly.
+ * Creates the research mode hook.
+ * Only injects on the first message of a session (messages array empty).
+ * Subsequent messages retain it via conversation history.
  */
 export function createResearchModeHook() {
-    return (_event, _context) => {
-        // Unconditionally inject research mode prompt
-        // The agent will decide whether to use skills based on task type
-        return {
-            prependContext: RESEARCH_MODE_PROMPT,
-        };
+    return (event, _context) => {
+        if (!event.messages || event.messages.length === 0) {
+            return { prependContext: RESEARCH_MODE_PROMPT };
+        }
+        return {};
     };
 }
 //# sourceMappingURL=research-mode.js.map

package/dist/src/hooks/research-mode.js.map

@@ -1 +1 @@
-{"version":3,"file":"research-mode.js","sourceRoot":"","sources":["../../../src/hooks/research-mode.ts"],"names":[],"mappings":"AAmBA;;;;;GAKG;AACH,MAAM,oBAAoB,GAAG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAiH5B,CAAC;AAEF;;;;;;GAMG;AACH,MAAM,UAAU,sBAAsB;IACpC,OAAO,CAAC,MAAiB,EAAE,QAAqB,EAAc,EAAE;QAC9D,8CAA8C;QAC9C,iEAAiE;QACjE,OAAO;YACL,cAAc,EAAE,oBAAoB;SACrC,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
+{"version":3,"file":"research-mode.js","sourceRoot":"","sources":["../../../src/hooks/research-mode.ts"],"names":[],"mappings":"AAmBA;;;;GAIG;AACH,MAAM,oBAAoB,GAAG,+KAA+K,CAAC;AAE7M;;;;GAIG;AACH,MAAM,UAAU,sBAAsB;IACpC,OAAO,CAAC,KAAgB,EAAE,QAAqB,EAAc,EAAE;QAC7D,IAAI,CAAC,KAAK,CAAC,QAAQ,IAAI,KAAK,CAAC,QAAQ,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACnD,OAAO,EAAE,cAAc,EAAE,oBAAoB,EAAE,CAAC;QAClD,CAAC;QACD,OAAO,EAAE,CAAC;IACZ,CAAC,CAAC;AACJ,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "scientify",
-  "version": "1.7.1",
+  "version": "1.7.3",
   "description": "Scientify - AI-powered research workflow automation for OpenClaw. Includes idea generation, literature review, research pipeline skills, and arxiv tool.",
   "type": "module",
   "main": "dist/index.js",

package/skills/install-scientify/SKILL.md

@@ -31,31 +31,37 @@ metadata:
 
 | Skill | Description |
 |-------|-------------|
-| **research-pipeline** | Orchestrator for end-to-end ML research. Spawns sub-agents for each phase. |
-| **research-survey** | Deep analysis of papers: extract formulas, produce method comparison. |
-| **research-plan** | 4-part implementation plan (Dataset/Model/Training/Testing). |
-| **research-implement** | Implement ML code, run 2-epoch validation with `uv` venv isolation. |
-| **research-review** | Review implementation against plan. Iterates up to 3 times. |
-| **research-experiment** | Full training + ablation experiments. |
-| **literature-survey** | Literature survey: search → filter → download → cluster → report. |
-| **idea-generation** | Generate research ideas from arXiv/GitHub papers. |
+| **research-pipeline** | End-to-end orchestrator. Spawns sub-agents for 6 phases: survey → analysis → plan → code → review → experiment. |
+| **literature-survey** | Search arXiv + OpenAlex → filter → download .tex sources → cluster → generate survey report. |
+| **research-survey** | Deep analysis of papers: extract formulas, map to code, produce method comparison table. |
+| **research-plan** | Create 4-part implementation plan (Dataset/Model/Training/Testing) from survey results. |
+| **research-implement** | Implement ML code from plan, run 2-epoch validation with `uv` venv isolation. |
+| **research-review** | Review implementation. Iterates fix → rerun → review up to 3 times. |
+| **research-experiment** | Full training + ablation experiments. Requires review PASS. |
+| **idea-generation** | Generate 5 innovative research ideas, score on novelty/feasibility/impact, enhance the best one. |
+| **write-review-paper** | Draft a review/survey paper from project research outputs. |
 
 ### Commands (Direct, no LLM)
 
 | Command | Description |
 |---------|-------------|
-| `/research-status` | Show workspace status |
-| `/papers` | List downloaded papers |
+| `/research-status` | Show workspace status and active project |
+| `/papers` | List downloaded papers with metadata |
 | `/ideas` | List generated ideas |
 | `/projects` | List all projects |
-| `/project-switch <id>` | Switch project |
-| `/project-delete <id>` | Delete project |
+| `/project-switch <id>` | Switch active project |
+| `/project-delete <id>` | Delete a project |
 
 ### Tools
 
-- **arxiv_search** - Search arXiv.org API for papers (metadata only)
-- **arxiv_download** - Download arXiv papers (.tex source or PDF)
-- **github_search** - Search GitHub repositories
+| Tool | Description |
+|------|-------------|
+| `arxiv_search` | Search arXiv papers. Returns metadata (title, authors, abstract, ID). Supports sorting by relevance/date. |
+| `arxiv_download` | Batch download papers by arXiv ID. Prefers .tex source (PDF fallback). |
+| `openalex_search` | Search cross-disciplinary papers via OpenAlex API. Returns DOI, authors, citation count, OA status. |
+| `unpaywall_download` | Download open access PDFs by DOI via Unpaywall. Non-OA papers silently skipped. |
+| `github_search` | Search GitHub repositories. Returns name, description, stars, URL. Supports language filtering. |
+| `paper_browser` | Paginated browsing of large paper files (.tex/.md) to avoid context overflow. |
 
 ## Installation
 
@@ -69,16 +75,22 @@ Or let OpenClaw install it automatically when you use this skill.
 
 ## Usage Examples
 
+### End-to-End Research
+
+```
+Research scaling laws for classical ML classifiers on Fashion-MNIST
+```
+
 ### Generate Research Ideas
 
 ```
-帮我调研 "长文档摘要" 领域，生成一些创新的研究想法
+Explore recent advances in protein folding and generate innovative research ideas
 ```
 
-### Daily Literature Tracking
+### Literature Survey Only
 
 ```
-帮我设置一个定时任务，每天检查 arXiv 上关于 "transformer efficiency" 的新论文，发到飞书
+Survey the latest papers on vision-language models for medical imaging
 ```
 
 ### Check Workspace
@@ -90,5 +102,5 @@ Or let OpenClaw install it automatically when you use this skill.
 ## Links
 
 - npm: https://www.npmjs.com/package/scientify
-- GitHub: https://github.com/tsingyuai/scientific
+- GitHub: https://github.com/tsingyuai/scientify
 - Author: tsingyuai

package/skills/research-pipeline/SKILL.md

@@ -79,7 +79,7 @@ metadata:
 | `task` | string | 是 | 子 agent 的完整任务描述 |
 | `label` | string | 否 | 显示标签（如 "Deep Survey"） |
 | `model` | string | 否 | 模型覆盖（如 `tsingyu/gemini-3-flash-preview`） |
-| `runTimeoutSeconds` | number | 否 | 超时秒数 |
+| `runTimeoutSeconds` | number | 否 | 超时秒数（**必须设置，推荐 1800**） |
 
 **task 字段格式**（子 agent 是独立 session，看不到当前上下文）：
 
@@ -124,6 +124,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果缺失，调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/literature-survey\n工作目录: {$W绝对路径}\n研究主题: {从task.json提取}\n请搜索、筛选、下载论文到工作目录的 papers/ 下。"`
 - label: `"Literature Survey"`
+- runTimeoutSeconds: `1800`
 
 **验证:** `ls $W/papers/_meta/*.json` 至少有 3 个文件
 
@@ -136,6 +137,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果缺失，先读取 Phase 1 摘要（论文数量、方向），然后调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/research-survey\n工作目录: {$W绝对路径}\n上下文: 已下载 {N} 篇论文，方向包括 {directions}。\n重点论文: {top 3 arxiv_id 和标题}\n请深度分析论文、提取公式，写入 survey_res.md。"`
 - label: `"Deep Survey"`
+- runTimeoutSeconds: `1800`
 
 **验证:** `$W/survey_res.md` 存在且包含"核心方法对比"表格
 
@@ -148,6 +150,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果缺失，读取 survey_res.md 摘要，然后调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/research-plan\n工作目录: {$W绝对路径}\n上下文: 调研发现核心方法是 {method}，推荐技术路线 {route}。\n关键公式: {1-2个公式}\n请制定实现计划到 plan_res.md。"`
 - label: `"Research Plan"`
+- runTimeoutSeconds: `1800`
 
 **验证:** `$W/plan_res.md` 存在且包含 4 个 section（Dataset/Model/Training/Testing）
 
@@ -160,6 +163,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果缺失，读取 plan_res.md 要点，然后调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/research-implement\n工作目录: {$W绝对路径}\n上下文:\n- 计划包含 {N} 个组件: {list}\n- 数据集: {dataset}\n- 框架: PyTorch\n请实现代码到 project/，运行 2 epoch 验证，写入 ml_res.md。"`
 - label: `"Research Implement"`
+- runTimeoutSeconds: `1800`
 
 **验证:**
 - `$W/project/run.py` 存在
@@ -175,6 +179,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果没有 PASS，调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/research-review\n工作目录: {$W绝对路径}\n上下文:\n- ml_res.md 显示 train_loss={value}\n- 计划在 plan_res.md\n请审查代码，如需修改则迭代修复（最多 3 轮）。"`
 - label: `"Research Review"`
+- runTimeoutSeconds: `1800`
 
 **验证:** 最新 `judge_v*.md` 中 `verdict: PASS` 或 `verdict: BLOCKED`
 
@@ -189,6 +194,7 @@ ACTIVE=$(cat ~/.openclaw/workspace/projects/.active 2>/dev/null)
 **如果缺失，调用 sessions_spawn 工具（然后停止，等待完成通知）：**
 - task: `"/research-experiment\n工作目录: {$W绝对路径}\n上下文:\n- Review PASS，代码已验证\n- plan_res.md 中指定 full epochs\n请执行完整训练 + 消融实验，写入 experiment_res.md。"`
 - label: `"Research Experiment"`
+- runTimeoutSeconds: `1800`
 
 **验证:** `$W/experiment_res.md` 包含 `[RESULT]` 行和消融表格