feat: IMAP mail sync with thread-to-contact matching (PRD §6.1)

Real IMAP connection via imapflow library:
- Connects to IMAP server (IMAPFlow, TLS on port 993)
- Fetches recent messages (configurable limit, default 20)
- Extracts: from, fromEmail, subject, date, snippet (200 chars)
- Thread-to-contact matching (PRD §6.1):
  1. Match sender email against persons table
  2. If no person match, extract domain and match against company.website
  3. Returns matchedCompany + matchedPerson for each email
- Audit-logged (action: crm.mailsync:imap)
- Zod validated (host, port, user, password, folder, limit)
- 502 on IMAP connection failure with detail
- 400 on validation failure

Tested: invalid creds → 502 'IMAP connection failed', missing host → 400

skills/aminer-daily-paper/README.md

@@ -0,0 +1,42 @@
+# aminer-daily-paper
+
+Personalized academic paper recommendation via the AMiner rec5 API. Intended for OpenClaw or similar hosts: the scripts return JSON with Markdown in `reply_text`; the host sends or displays that content.
+
+## Requirements
+
+- Python 3.10+
+- `AMINER_API_KEY` environment variable — obtain at https://open.aminer.cn/open/board?tab=control
+
+## Installation
+
+```bash
+pip install -r requirements.txt
+```
+
+## Usage
+
+```
+/aminer-dp
+/aminer-dp topics: multimodal agents, tool-use
+/aminer-dp scholar: Jie Tang org: Tsinghua
+/aminer-dp aminer_author_id: 696259801cb939bc391d3a37 topics: RAG, LLM
+recommend me recent papers on multimodal agents
+```
+
+The model (running the skill) extracts `topics`, `author_name`, `author_org`, or `aminer_author_id` from natural language input before calling the script.
+
+## How It Works
+
+1. `handle_trigger.py` parses the trigger text and runs `run_pipeline.py` as a subprocess.
+2. `run_pipeline.py` calls the AMiner rec5 API and writes `papers_summarized.json` under the output directory.
+3. The pipeline returns `final_response: "TEXT"` and `reply_text` (Markdown). No Feishu card builders or `openclaw` dispatch live in this skill.
+
+## API
+
+```
+POST https://publicapi.chatglm.cn/chatglm_public/skill/aminer/api/v3/paper/rec5
+Authorization: <AMINER_API_KEY>
+```
+
+Key request fields: `aminer_author_id`, `author_name`, `author_org`, `topics`, `size`, `language_sort`.  
+At least one of `aminer_author_id`, `author_name`, or `topics` should be provided. When none are given, the API returns personalized recommendations via the token-linked account.

skills/aminer-daily-paper/SKILL.md

@@ -0,0 +1,212 @@
+---
+name: aminer-daily-paper
+version: 1.1.2
+description: "Personalized academic paper recommendation via AMiner rec5 API. Activate this skill whenever the user asks for paper recommendations, whether triggered by /aminer-dp, /skill aminer-dp, or any natural language request such as 'recommend me papers on multimodal agents'. When invoked: extract topics/scholar signals from the input yourself, call handle_trigger.py with structured fields, then present the Markdown in `reply_text` to the user."
+user-invocable: true
+disable-model-invocation: false
+metadata:
+  {
+    "openclaw":
+      {
+        "emoji": "📚",
+        "requires": {
+          "bins": ["python3"],
+          "env": ["AMINER_API_KEY"]
+        },
+        "primaryEnv": "AMINER_API_KEY"
+      }
+  }
+---
+
+# aminer-daily-paper
+
+Personalized paper recommendation via AMiner rec5 API. Token required: set `AMINER_API_KEY` env var.
+- Docs: https://open.aminer.cn/open/docs | Console: https://open.aminer.cn/open/board?tab=control
+
+**When to activate**: any time the user asks for paper recommendations — explicit command (`/aminer-dp ...`) or natural language (`recommend me papers on RAG`, `帮我推荐最近的多模态论文`).
+
+---
+
+## Pre-flight: Check Required Environment Variables
+
+**`AMINER_API_KEY`** — Always required. Check before calling the script:
+
+```bash
+[ -z "${AMINER_API_KEY+x}" ] && echo "AMINER_API_KEY missing" || echo "AMINER_API_KEY exists"
+```
+
+If missing, stop and tell the user:
+> `AMINER_API_KEY` is not set. Please obtain a token at https://open.aminer.cn and set it as an environment variable.
+
+No other environment variables are required.
+
+---
+
+## API Endpoint
+
+```
+POST https://publicapi.chatglm.cn/chatglm_public/skill/aminer/api/v3/paper/rec5
+Authorization: ${AMINER_API_KEY}
+Content-Type: application/json;charset=utf-8
+```
+
+### Request Fields
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `author_name` | string | conditional | Scholar name (English). The backend resolves it to a scholar ID via person search. |
+| `author_org` | string | optional | Scholar institution (English full name). Required for disambiguation when the name is ambiguous. |
+| `topics` | string[] | conditional | Research topic phrases. **Use the user’s wording** (Chinese, English, or mixed). The API accepts multi-language topic strings. |
+| `size` | int | optional | Number of papers per call (1–20). Omit to let the model decide (see below). |
+| `offset` | int | optional | Pagination offset (0–100, default 0). |
+| `language_sort` | string | optional | `zh` or `en` **only when the user explicitly asks** for Chinese- or English-biased ranking (e.g. “优先中文论文” / “prefer English papers”). Otherwise omit; the request will not include this field. |
+
+At least one of `author_name` or `topics` should be provided. When none are given, the API returns personalized recommendations based on the account associated with `AMINER_API_KEY`.
+
+### Response Structure
+
+```json
+{
+  "code": 200,
+  "success": true,
+  "data": [{
+    "offset": 0,
+    "size": 5,
+    "total": 32,
+    "papers": [{
+      "paper_id": "...",
+      "arxiv_id": "",
+      "title": "...",
+      "year": 2026,
+      "authors": ["Author A", "Author B"],
+      "keywords": ["kw1", "kw2"],
+      "summary": "...",
+      "structured_summary": {
+        "research_problem": "...",
+        "research_challenge": "...",
+        "research_method": "...",
+        "experimental_results": ""
+      },
+      "famous_authors": [],
+      "aminer_author_profiles": [],
+      "author_entries": [],
+      "links": {
+        "aminer": "https://www.aminer.cn/pub/{paper_id}",
+        "arxiv": "",
+        "pdf": ""
+      },
+      "paper_url": "https://www.aminer.cn/pub/{paper_id}",
+      "source": "local_rec5"
+    }]
+  }]
+}
+```
+
+---
+
+## Input Formats
+
+Structured commands or plain natural language — both are valid.
+
+```
+/aminer-dp
+/aminer-dp topics: multimodal agents, tool-use
+/aminer-dp scholar: Jie Tang org: Tsinghua papers: OAG-Bench | RPC-Bench
+recommend me recent papers on RAG
+```
+
+`/aminer-dp` with no parameters calls the API with only the token — the API uses `AMINER_API_KEY` to identify the account and returns personalized recommendations.
+
+**Natural language input** — you (the model) must parse it into fields before calling the script. **Critical for `topics`:**
+
+1. **`topics` — do not “translate away” the user’s intent**
+   - If the user already wrote `topics:` in the trigger (e.g. `具身智能`, `环境保护`), pass **those exact strings** into `handle_trigger.py`’s `--text`. **Do not** replace them with unrelated English terms (e.g. do **not** map arbitrary topics to “Knowledge Distillation”, “Smart agriculture”, or any other field the user did not ask for).
+   - If you add English for retrieval, it must be a **faithful** alias of the same concept (e.g. 具身智能 → `embodied intelligence`, 环境保护 → `environmental protection`). When in doubt, **keep the user’s original words** and do not invent synonyms.
+   - **Never** change the user’s topic into a different research area.
+
+2. **Scholars and institutions (person search still English-oriented)**
+   - `author_name` / `author_org`: use commonly used **English** forms when resolving scholars (e.g. `Jie Tang`, `Tsinghua University`), expand well-known institution abbreviations to full official names, and add `author_org` when the name is ambiguous. If you cannot map a name safely, ask the user.
+
+3. **`language_sort`** — Put `language_sort: zh` or `language_sort: en` in the trigger **only if** the user clearly wants recommendations ranked with a **Chinese** or **English** preference. If they did not ask, **do not** add it (the API call omits `language_sort`).
+
+4. Decide `size` and whether to make multiple calls (see **Call Strategy**).
+5. Reconstruct the trigger, then call `handle_trigger.py`.
+
+Example (Chinese topics — **keep as-is**):
+- User: `/aminer-dp topics: 具身智能, 环境保护`
+- You call: `handle_trigger.py --text "/aminer-dp topics: 具身智能, 环境保护"`  
+  (Do **not** rewrite `topics` into unrelated English.)
+
+Example:
+- User: `/aminer-dp 我做多模态智能体和 tool-use，帮我推荐最近论文`
+- You extract: `topics: multimodal agents, tool-use`
+- You call: `handle_trigger.py --text "/aminer-dp topics: multimodal agents, tool-use size: 5"`
+
+Example (scholar):
+- User: `/aminer-dp 我是唐杰，清华大学，做多模态和知识图谱`
+- You extract: `scholar: Jie Tang, org: Tsinghua University, topics: multimodal, knowledge graph`
+- You call: `handle_trigger.py --text "/aminer-dp scholar: Jie Tang org: Tsinghua University topics: multimodal, knowledge graph"`
+
+Example (ambiguous name, ask user):
+- User: `/aminer-dp 推荐张伟方向的论文`
+- You: "张伟是一个常见名字，请提供机构信息以便精确匹配，例如：张伟，北京大学。或者直接提供 aminer_author_id。"
+
+**`papers` field**: representative paper titles (e.g. `papers: OAG-Bench | RPC-Bench`) accompany `scholar`/`author_name` for disambiguation context. They do not map directly to an API field.
+
+---
+
+## Call Strategy
+
+You decide `size` and whether to make multiple calls based on the input:
+
+| Scenario | Action |
+|----------|--------|
+| Single topic or scholar, casual request | 1 call, omit `size` (default 10) |
+| User explicitly asks for a number (e.g. "give me 5") | 1 call, honor the number (max 20) |
+| Multiple distinct topics (e.g. RAG + multimodal agents) | 1 call per topic group, `size: 5` each |
+| Broad open-ended request with no topics | 1 call, omit `size` (default 10) |
+
+**Multi-call rules:**
+- Call `handle_trigger.py` once per topic group, passing a focused `topics:` subset each time.
+- Keep each `topics:` list to 1–3 closely related terms for precision.
+- Make calls sequentially; present all results together after all calls finish.
+- Total papers across all calls should not exceed ~15 unless the user asks for more.
+
+---
+
+## Execution
+
+Only one supported entrypoint:
+
+```bash
+python3 "{baseDir}/scripts/handle_trigger.py" \
+  --base-dir "{baseDir}" \
+  --text "<trigger text with explicit fields>" \
+  [--config /path/to/config.yaml]
+```
+
+- `--text`: reconstructed trigger with explicit fields (`topics:`, `scholar:`, etc.)
+- `--config`: optional path to a YAML config (defaults to `{baseDir}/config.yaml` when the file exists, via the runtime copy under `outputs/`)
+
+`handle_trigger.py` parses the fields, calls the rec5 API, and returns JSON including `reply_text` (Markdown) for you to show to the user.
+
+---
+
+## Contract
+
+- Every explicit invocation is a new run.
+- Do not answer with status-only text.
+- Do not search, install, or repair skills.
+- After running `handle_trigger.py`, check `final_response` in the JSON output:
+  - `TEXT` — Normal path. Present `reply_text` (Markdown) to the user. Optional: you may still refine wording for the active channel; `prompts/enrich.md` is a reference for Chinese enrichment if you want richer copy.
+  - Any error → report the `reply_text` (or error detail) to the user.
+
+**Note:** The skill only returns JSON with `reply_text`; it does not implement channel-specific sending.
+
+---
+
+## Error Handling
+
+- `AMINER_API_KEY` missing → stop, prompt user to set it.
+- No profile input → prompt user to provide topics, scholar name, or `aminer_author_id`.
+- API error → report the error stage; do not fall back to other skills.

skills/aminer-daily-paper/commands/aminer-dp.md

@@ -0,0 +1,52 @@
+---
+description: AMiner 个性化论文推荐 — 按主题或学者画像
+argument-hint: [topics: 主题1, 主题2 | scholar: 姓名 org: 机构 | 自然语言]
+allowed-tools: Read, Bash, Glob, Grep
+---
+
+# /aminer-dp — AMiner Daily Paper
+
+User invoked the AMiner daily paper recommendation skill with the following arguments:
+
+```
+$ARGUMENTS
+```
+
+## Your task
+
+Strictly follow `${CLAUDE_PLUGIN_ROOT}/SKILL.md` (English) or `${CLAUDE_PLUGIN_ROOT}/SKILL.zh.md` (Chinese). Key rules summarized below — but read the SKILL file if you need detail.
+
+### 1. Pre-flight
+
+Verify the `AMINER_API_KEY` env var is set:
+
+```bash
+[ -z "${AMINER_API_KEY+x}" ] && echo "AMINER_API_KEY missing" || echo "AMINER_API_KEY exists"
+```
+
+If missing, stop and tell the user to set it (token from <https://open.aminer.cn>). Do not call the script.
+
+### 2. Parse `$ARGUMENTS` into structured fields
+
+Extract any of: `topics`, `scholar` / `author_name`, `org` / `author_org`, `papers`, `size`, `language_sort`.
+
+**Critical rules — do NOT violate:**
+
+- **`topics`**: keep the user's exact wording. If they wrote `具身智能, 环境保护`, pass those Chinese strings through. **Never** translate them into unrelated English fields (e.g. don't map them to "Knowledge Distillation" or "Smart agriculture"). If you add an English alias, it must be a faithful translation of the same concept (e.g. 具身智能 → embodied intelligence).
+- **`language_sort`**: include `zh` or `en` **only when the user explicitly asks** for Chinese-/English-biased ranking (e.g. "优先中文论文" / "prefer English papers"). Otherwise omit it entirely.
+- **`scholar` / `org`**: use English-canonical names where reasonable (e.g. `Jie Tang`, `Tsinghua University`). If a Chinese name is ambiguous and no org is given, ask the user before guessing.
+- If `$ARGUMENTS` is empty, call the script with no extra fields — the API will return personalized recs based on the API key.
+
+### 3. Run the entrypoint
+
+Reconstruct the trigger and execute (one call per topic group; ≤3 closely related terms per group):
+
+```bash
+python3 "${CLAUDE_PLUGIN_ROOT}/scripts/handle_trigger.py" \
+  --base-dir "${CLAUDE_PLUGIN_ROOT}" \
+  --text "/aminer-dp <reconstructed fields>"
+```
+
+### 4. Present the result
+
+Parse the JSON output. When `final_response == "TEXT"`, render the `reply_text` (Markdown) directly to the user. On error, surface the error message; do not fall back to other skills.

skills/aminer-daily-paper/prompts/enrich.md

@@ -0,0 +1,40 @@
+# Paper Enrichment Prompt
+
+You are enriching academic papers with Chinese content before they are sent as Feishu cards.
+
+For each paper in the `papers` array that is missing a Chinese summary (i.e., the `summary` field contains only English text or is empty), generate:
+
+1. **summary** — A concise Chinese summary (1-2 sentences, starting with "本文"). Describe the core contribution and significance. Keep English terms (e.g., RAG, LLM, Transformer) as-is.
+2. **keywords** — 2-4 Chinese keywords. Keep well-known English acronyms as-is.
+3. **comment** — If the paper's venue is a well-known conference/journal, annotate its tier, e.g., "已发表在 AAAI（CCF-A）". Leave empty if unknown.
+
+Rules:
+- Do NOT fabricate information not present in the paper's title/abstract.
+- Do NOT generate or modify `famous_authors`.
+- If the abstract is missing or too short to summarize, write "摘要信息不足，请查看原文。"
+- Preserve all other fields unchanged.
+
+## Input
+
+The papers are in `papers_summarized.json` at the path provided. The file structure:
+
+```json
+{
+  "status": "success",
+  "profile_topics": ["topic1", "topic2"],
+  "papers": [
+    {
+      "title": "...",
+      "summary": "...(may be English-only)...",
+      "keywords": ["kw1", "kw2"],
+      "authors": ["..."],
+      "comment": "",
+      ...other fields preserved as-is...
+    }
+  ]
+}
+```
+
+## Output
+
+Write the enriched data back to the same `papers_summarized.json` path, preserving all fields. Only update `summary`, `keywords`, and `comment` for papers that need enrichment.

skills/aminer-daily-paper/requirements.txt

@@ -0,0 +1 @@
+PyYAML>=6.0

skills/aminer-daily-paper/scripts/__init__.py

@@ -0,0 +1 @@
+"""aminer_rec pipeline package."""

skills/aminer-daily-paper/scripts/common.py

@@ -0,0 +1,28 @@
+from __future__ import annotations
+
+import json
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+
+
+def utc_now_iso() -> str:
+    return datetime.now(timezone.utc).replace(microsecond=0).isoformat()
+
+
+def read_json(path: Path) -> dict[str, Any]:
+    return json.loads(path.read_text(encoding="utf-8"))
+
+
+def write_json(path: Path, data: dict[str, Any]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(
+        json.dumps(
+            data,
+            ensure_ascii=False,
+            indent=2,
+            default=lambda value: value.isoformat() if isinstance(value, datetime) else str(value),
+        )
+        + "\n",
+        encoding="utf-8",
+    )

skills/aminer-daily-paper/scripts/constants.py

@@ -0,0 +1,7 @@
+from __future__ import annotations
+
+DEFAULT_TOP_K = 10
+
+AMINER_AUTHOR_URL_TEMPLATE = "https://www.aminer.cn/profile/{author_id}"
+AMINER_PAPER_URL_TEMPLATE = "https://www.aminer.cn/pub/{paper_id}"
+DEFAULT_REC5_URL = "https://publicapi.chatglm.cn/chatglm_public/skill/aminer/api/v3/paper/rec5"

skills/aminer-daily-paper/scripts/handle_trigger.py

@@ -0,0 +1,527 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+if __package__ in {None, ""}:
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+def _clean_text(value: Any) -> str:
+    return re.sub(r"\s+", " ", str(value or "")).strip()
+
+
+def _split_topics(text: str) -> list[str]:
+    pieces = re.split(r"[,，;/；、\n]+", text)
+    topics: list[str] = []
+    for piece in pieces:
+        topic = _clean_text(piece)
+        if topic and topic not in topics:
+            topics.append(topic)
+    return topics
+
+
+def _split_papers(text: str) -> list[str]:
+    pieces = re.split(r"[|\n;；]+", text)
+    papers: list[str] = []
+    for piece in pieces:
+        paper = _clean_text(piece)
+        if paper and paper not in papers:
+            papers.append(paper)
+    return papers
+
+
+def _extract_command_text(raw_text: str) -> str:
+    lines = [line.strip() for line in str(raw_text or "").splitlines() if line.strip()]
+    for line in reversed(lines):
+        match = re.search(r"(/(?:skill\s+)?aminer[-_]dp\b.*)$", line, flags=re.IGNORECASE)
+        if match:
+            return match.group(1).strip()
+    return str(raw_text or "")
+
+
+FIELD_LABELS = {
+    "aminer_author_id": ["aminer_author_id"],
+    "topics": ["topics", "topic", "方向", "研究方向"],
+    "scholar_name": ["scholar", "name", "author", "学者", "作者"],
+    "scholar_org": ["org", "organization", "affiliation", "机构", "单位"],
+    "paper_titles": ["paper", "papers", "代表作", "论文"],
+    "papers_file": ["papers_file", "source_file", "profile_file", "文件", "路径"],
+    "language_sort": ["language_sort"],
+    "size": ["size"],
+}
+
+
+GENERIC_REQUEST_PATTERNS = [
+    r"帮我推荐(?:一下)?论文",
+    r"推荐(?:一下)?论文",
+    r"推荐一些论文",
+    r"给我推荐(?:一下)?论文",
+    r"推荐最近论文",
+    r"想看论文",
+]
+
+ORG_HINT_PATTERNS = (
+    r"大学",
+    r"学院",
+    r"研究院",
+    r"研究所",
+    r"实验室",
+    r"中心",
+    r"University",
+    r"College",
+    r"Institute",
+    r"Laboratory",
+    r"Lab\b",
+    r"School",
+    r"Department",
+)
+
+
+MAX_TOPICS = 8
+MAX_TOPIC_LENGTH = 80
+MAX_PAPER_TITLES = 8
+MAX_PAPER_TITLE_LENGTH = 300
+MAX_SCHOLAR_NAME_LENGTH = 80
+MAX_SCHOLAR_ORG_LENGTH = 160
+MAX_FREE_TEXT_LENGTH = 600
+ALLOWED_PAPERS_FILE_SUFFIXES = {".json"}
+TOPIC_STOPWORDS = {
+    "论文",
+    "推荐",
+    "一下",
+    "推荐一下",
+    "相关论文",
+    "papers",
+    "paper",
+    "recommend",
+    "recommendation",
+    "research papers",
+}
+TOPIC_STOPWORDS_CASEFOLD = {item.casefold() for item in TOPIC_STOPWORDS}
+
+
+def _capture_field(command_body: str, field_name: str) -> str:
+    labels = FIELD_LABELS[field_name]
+    all_labels = [re.escape(label) for values in FIELD_LABELS.values() for label in values]
+    pattern = rf"(?:{'|'.join(re.escape(label) for label in labels)})\s*[:：]\s*(.+?)(?=\s*(?:{'|'.join(all_labels)})\s*[:：]|$)"
+    match = re.search(pattern, command_body, flags=re.IGNORECASE | re.S)
+    return _clean_text(match.group(1)) if match else ""
+
+
+def _truncate_text(value: Any, max_length: int) -> str:
+    cleaned = _clean_text(value)
+    if len(cleaned) <= max_length:
+        return cleaned
+    return cleaned[:max_length].strip()
+
+
+def _normalize_topics_for_interface(values: list[Any]) -> list[str]:
+    topics: list[str] = []
+    for value in list(values or []):
+        topic = _truncate_text(value, MAX_TOPIC_LENGTH)
+        if topic and topic not in topics:
+            topics.append(topic)
+        if len(topics) >= MAX_TOPICS:
+            break
+    return topics
+
+
+def _normalize_paper_titles_for_interface(values: list[Any]) -> list[str]:
+    paper_titles: list[str] = []
+    for value in list(values or []):
+        paper_title = _truncate_text(value, MAX_PAPER_TITLE_LENGTH)
+        if paper_title and paper_title not in paper_titles:
+            paper_titles.append(paper_title)
+        if len(paper_titles) >= MAX_PAPER_TITLES:
+            break
+    return paper_titles
+
+
+def _resolve_interface_papers_file(base_dir: Path, path_text: str) -> str:
+    cleaned = _clean_text(path_text)
+    if not cleaned:
+        return ""
+
+    candidate = Path(cleaned).expanduser()
+    resolved_base_dir = base_dir.resolve()
+    resolved_candidate = (resolved_base_dir / candidate).resolve() if not candidate.is_absolute() else candidate.resolve()
+    try:
+        resolved_candidate.relative_to(resolved_base_dir)
+    except ValueError as exc:
+        raise ValueError("papers_file_outside_base_dir") from exc
+
+    if resolved_candidate.suffix.lower() not in ALLOWED_PAPERS_FILE_SUFFIXES:
+        raise ValueError("unsupported_papers_file")
+    return str(resolved_candidate)
+
+
+def _normalize_interface_payload(parsed: dict[str, Any], *, base_dir: Path) -> dict[str, Any]:
+    normalized = dict(parsed)
+    raw_uid = _clean_text(parsed.get("raw_aminer_author_id"))
+    if raw_uid and not re.fullmatch(r"[0-9a-fA-F]{24}", raw_uid):
+        raise ValueError("invalid_aminer_author_id")
+
+    normalized["aminer_author_id"] = _clean_text(parsed.get("aminer_author_id"))
+    normalized["topics"] = _normalize_topics_for_interface(list(parsed.get("topics") or []))
+    normalized["scholar_name"] = _truncate_text(parsed.get("scholar_name"), MAX_SCHOLAR_NAME_LENGTH)
+    normalized["scholar_org"] = _truncate_text(parsed.get("scholar_org"), MAX_SCHOLAR_ORG_LENGTH)
+    normalized["paper_titles"] = _normalize_paper_titles_for_interface(list(parsed.get("paper_titles") or []))
+    normalized["papers_file"] = _resolve_interface_papers_file(base_dir, str(parsed.get("papers_file") or ""))
+    normalized["free_text"] = _truncate_text(parsed.get("free_text"), MAX_FREE_TEXT_LENGTH)
+    lang = _clean_text(parsed.get("language_sort"))
+    normalized["language_sort"] = lang if lang in {"zh", "en"} else ""
+    raw_size = _clean_text(parsed.get("size"))
+    try:
+        normalized["size"] = max(1, min(int(raw_size), 20)) if raw_size else 0
+    except (ValueError, TypeError):
+        normalized["size"] = 0
+    if not normalized["topics"] and normalized["free_text"]:
+        normalized["topics"] = _infer_topics_from_free_text(normalized["free_text"])
+    return normalized
+
+
+def _strip_explicit_fields(command_body: str) -> str:
+    all_labels = [re.escape(label) for values in FIELD_LABELS.values() for label in values]
+    pattern = rf"(?:{'|'.join(all_labels)})\s*[:：]\s*.+?(?=\s*(?:{'|'.join(all_labels)})\s*[:：]|$)"
+    cleaned = re.sub(pattern, " ", command_body, flags=re.IGNORECASE | re.S)
+    return _clean_text(cleaned)
+
+
+def _remove_generic_request_phrases(text: str) -> str:
+    cleaned = str(text or "")
+    for pattern in GENERIC_REQUEST_PATTERNS:
+        cleaned = re.sub(pattern, " ", cleaned, flags=re.IGNORECASE)
+    cleaned = re.sub(r"[。！？!?.，,；;、\s]+", " ", cleaned)
+    return _clean_text(cleaned)
+
+
+def _normalize_topic_candidate(text: str) -> str:
+    candidate = _clean_text(text)
+    if not candidate:
+        return ""
+    candidate = re.sub(r'^[\'"`“”‘’]+|[\'"`“”‘’]+$', "", candidate)
+    candidate = re.sub(r"^(?:关于|研究(?:方向)?|方向|领域|做|关注|topic(?:s)?|about|on)\s*", "", candidate, flags=re.IGNORECASE)
+    candidate = re.sub(r"\s*(?:相关|方向|领域|论文|papers?|research)\s*$", "", candidate, flags=re.IGNORECASE)
+    candidate = _clean_text(candidate)
+    if not candidate:
+        return ""
+    if candidate.casefold() in TOPIC_STOPWORDS_CASEFOLD:
+        return ""
+    return _truncate_text(candidate, MAX_TOPIC_LENGTH)
+
+
+def _infer_topics_from_free_text(text: str) -> list[str]:
+    cleaned = _clean_text(text)
+    if not cleaned:
+        return []
+
+    for pattern in GENERIC_REQUEST_PATTERNS:
+        cleaned = re.sub(pattern, " ", cleaned, flags=re.IGNORECASE)
+    cleaned = re.sub(r"\b(?:recommend|papers?|please|find|show)\b", " ", cleaned, flags=re.IGNORECASE)
+    cleaned = re.sub(r"(?:给我|帮我|请|想看|想要|推荐|看看)\s*", " ", cleaned)
+    cleaned = _clean_text(cleaned)
+    if not cleaned:
+        return []
+
+    parts = re.split(r"[,，;/；、\n]+|\s+(?:and|or|以及|和|与|及)\s+", cleaned, flags=re.IGNORECASE)
+    topics: list[str] = []
+    seen: set[str] = set()
+    for part in parts:
+        candidate = _normalize_topic_candidate(part)
+        if not candidate:
+            continue
+        key = candidate.casefold()
+        if key in seen:
+            continue
+        seen.add(key)
+        topics.append(candidate)
+        if len(topics) >= MAX_TOPICS:
+            break
+    if topics:
+        return topics
+
+    fallback = _normalize_topic_candidate(cleaned)
+    return [fallback] if fallback else []
+
+
+def _infer_scholar_from_free_text(text: str) -> tuple[str, str, str]:
+    normalized = _clean_text(text)
+    if not normalized:
+        return "", "", ""
+
+    patterns = [
+        r"^我(?:是|叫)\s*(?P<name>[^，,。；;、\s]{2,20})\s*[，,、]\s*(?P<org>[^。；;，,]{2,60})",
+        r"^本人(?:是)?\s*(?P<name>[^，,。；;、\s]{2,20})\s*[，,、]\s*(?P<org>[^。；;，,]{2,60})",
+        r"^我是\s*(?P<org>[^，,。；;]{2,60})\s*的\s*(?P<name>[^，,。；;、\s]{2,20})",
+    ]
+    for pattern in patterns:
+        match = re.search(pattern, normalized, flags=re.IGNORECASE)
+        if not match:
+            continue
+        scholar_name = _clean_text(match.groupdict().get("name"))
+        scholar_org = _clean_text(match.groupdict().get("org"))
+        if not scholar_name:
+            continue
+        residual = _clean_text(normalized[match.end() :])
+        residual = _remove_generic_request_phrases(residual)
+        scholar_org = re.split(r"[。！？!?.；;]", scholar_org, maxsplit=1)[0].strip()
+        scholar_org = _remove_generic_request_phrases(scholar_org)
+        return scholar_name, scholar_org, residual
+
+    bare_match = re.search(
+        r"^(?P<name>[A-Za-z][A-Za-z .'-]{1,60}|[\u4e00-\u9fff·]{2,20})\s*[，,、]\s*(?P<org>[^。；;，,\n]{2,80})",
+        normalized,
+        flags=re.IGNORECASE,
+    )
+    if bare_match:
+        scholar_name = _clean_text(bare_match.group("name"))
+        scholar_org = _clean_text(bare_match.group("org"))
+        org_hint_pattern = "|".join(ORG_HINT_PATTERNS)
+        if scholar_name and scholar_org and re.search(org_hint_pattern, scholar_org, flags=re.IGNORECASE):
+            residual = _clean_text(normalized[bare_match.end() :])
+            residual = _remove_generic_request_phrases(residual)
+            scholar_org = re.split(r"[。！？!?.；;]", scholar_org, maxsplit=1)[0].strip()
+            scholar_org = _remove_generic_request_phrases(scholar_org)
+            return scholar_name, scholar_org, residual
+    return "", "", normalized
+
+
+def parse_trigger_text(text: str) -> dict[str, Any]:
+    raw_text = str(text or "")
+    command_text = _extract_command_text(raw_text)
+    normalized = _clean_text(command_text)
+    is_trigger = bool(re.search(r"^/(skill\s+)?aminer[-_]dp\b", normalized, flags=re.IGNORECASE))
+    body = re.sub(r"^/(skill\s+)?aminer[-_]dp\b", "", command_text, flags=re.IGNORECASE).strip()
+
+    uid_match = re.search(r"aminer_author_id\s*[:：]\s*([0-9a-fA-F]{24})", body, flags=re.IGNORECASE)
+    uid = uid_match.group(1) if uid_match else ""
+    scholar_name = _capture_field(body, "scholar_name")
+    scholar_org = _capture_field(body, "scholar_org")
+    free_text = _strip_explicit_fields(body)
+    if not scholar_name:
+        inferred_name, inferred_org, residual = _infer_scholar_from_free_text(free_text)
+        if inferred_name:
+            scholar_name = inferred_name
+            if inferred_org and not scholar_org:
+                scholar_org = inferred_org
+            free_text = residual
+
+    return {
+        "raw_text": raw_text,
+        "command_text": command_text,
+        "raw_aminer_author_id": _capture_field(body, "aminer_author_id"),
+        "aminer_author_id": uid,
+        "topics": _split_topics(_capture_field(body, "topics")),
+        "scholar_name": scholar_name,
+        "scholar_org": scholar_org,
+        "paper_titles": _split_papers(_capture_field(body, "paper_titles")),
+        "papers_file": _capture_field(body, "papers_file"),
+        "language_sort": _capture_field(body, "language_sort"),
+        "size": _capture_field(body, "size"),
+        "free_text": free_text,
+        "is_trigger": is_trigger,
+    }
+
+
+def _load_config(base_dir: Path, config_path: Path | None) -> dict[str, Any]:
+    if config_path and config_path.exists():
+        return yaml.safe_load(config_path.read_text(encoding="utf-8")) or {}
+    default_path = base_dir / "config.yaml"
+    if default_path.exists():
+        return yaml.safe_load(default_path.read_text(encoding="utf-8")) or {}
+    return {}
+
+
+def _run_pipeline(
+    *,
+    base_dir: Path,
+    output_dir: Path,
+    config_path: Path | None,
+    aminer_author_id: str,
+    topics: list[str],
+    scholar_name: str,
+    scholar_org: str,
+    paper_titles: list[str],
+    papers_file: str,
+    free_text: str,
+    language_sort: str,
+    size: int,
+) -> dict[str, Any]:
+    command = [
+        sys.executable,
+        str(base_dir / "scripts" / "run_pipeline.py"),
+        "--base-dir",
+        str(base_dir),
+        "--output-dir",
+        str(output_dir),
+    ]
+    if config_path is not None:
+        command.extend(["--config", str(config_path)])
+    if aminer_author_id.strip():
+        command.extend(["--aminer-author-id", aminer_author_id.strip()])
+    if language_sort.strip():
+        command.extend(["--language-sort", language_sort.strip()])
+    if size > 0:
+        command.extend(["--size", str(size)])
+    if topics:
+        command.extend(["--topics", *topics])
+    if scholar_name.strip():
+        command.extend(["--scholar-name", scholar_name.strip()])
+    if scholar_org.strip():
+        command.extend(["--scholar-org", scholar_org.strip()])
+    for paper_title in paper_titles:
+        if paper_title.strip():
+            command.extend(["--paper-title", paper_title.strip()])
+    if papers_file.strip():
+        command.extend(["--papers-file", papers_file.strip()])
+    if free_text.strip():
+        command.extend(["--free-text", free_text.strip()])
+    completed = subprocess.run(command, capture_output=True, text=True, check=False)
+    if completed.returncode != 0:
+        detail = completed.stderr.strip() or completed.stdout.strip() or "run_pipeline failed"
+        raise RuntimeError(detail)
+    return json.loads(completed.stdout)
+
+
+def _compact_pipeline_error(detail: str) -> str:
+    text = _clean_text(detail)
+    if not text:
+        return "unknown_error"
+    if "Traceback" not in text:
+        return text
+    lines = [line.strip() for line in str(detail or "").splitlines() if line.strip()]
+    for line in reversed(lines):
+        if line.startswith("RuntimeError:"):
+            return _clean_text(line.split("RuntimeError:", 1)[1])
+    return _clean_text(lines[-1]) if lines else text
+
+
+def handle_trigger(
+    *,
+    base_dir: Path,
+    text: str,
+    config_path: Path | None = None,
+) -> dict[str, Any]:
+    parsed = parse_trigger_text(text)
+    try:
+        parsed = _normalize_interface_payload(parsed, base_dir=base_dir)
+    except ValueError as exc:
+        detail = _clean_text(str(exc))
+        if detail == "invalid_aminer_author_id":
+            reply_text = "输入里的 `aminer_author_id` 不合法。请提供 24 位十六进制字符串，例如：`/aminer-dp aminer_author_id: 696259801cb939bc391d3a37 topics: 多模态, 智能体`。"
+        elif detail == "papers_file_outside_base_dir":
+            reply_text = "出于安全限制，`papers_file` 只能指向当前 skill 目录内的 JSON 文件，不能引用目录外路径。"
+        elif detail == "unsupported_papers_file":
+            reply_text = "`papers_file` 目前只支持 `.json` 文件。"
+        else:
+            reply_text = f"输入不符合接口约束：{detail}"
+        return {
+            "status": "success",
+            "mode": "invalid_input",
+            "final_response": "TEXT",
+            "reply_text": reply_text,
+        }
+    has_profile_input = bool(
+        parsed["aminer_author_id"]
+        or parsed["topics"]
+        or parsed["scholar_name"]
+        or parsed["scholar_org"]
+        or parsed["paper_titles"]
+        or parsed["papers_file"]
+        or parsed["free_text"]
+    )
+    if not parsed["is_trigger"] and not has_profile_input:
+        return {
+            "status": "success",
+            "mode": "help",
+            "final_response": "TEXT",
+            "reply_text": "请发送 `/aminer-dp`（直接推荐），或加上 `topics: 多模态, 智能体`、`scholar: Jie Tang` 等参数精确推荐，也可以直接描述研究方向。",
+        }
+
+    output_dir = base_dir / "outputs"
+    output_dir.mkdir(parents=True, exist_ok=True)
+    loaded_config = _load_config(base_dir, config_path)
+    runtime_config_path = output_dir / "runtime_config.yaml"
+    runtime_config_path.write_text(yaml.safe_dump(loaded_config, allow_unicode=True, sort_keys=False), encoding="utf-8")
+
+    try:
+        pipeline_result = _run_pipeline(
+            base_dir=base_dir,
+            output_dir=output_dir,
+            config_path=runtime_config_path,
+            aminer_author_id=parsed["aminer_author_id"],
+            topics=parsed["topics"],
+            scholar_name=parsed["scholar_name"],
+            scholar_org=parsed["scholar_org"],
+            paper_titles=parsed["paper_titles"],
+            papers_file=parsed["papers_file"],
+            free_text=parsed["free_text"],
+            language_sort=parsed["language_sort"],
+            size=parsed["size"],
+        )
+    except Exception as exc:
+        detail = _compact_pipeline_error(str(exc).strip())
+        if parsed["aminer_author_id"] and not parsed["topics"] and ("profile_unavailable" in detail or "missing_topics" in detail or "no_bind_papers_or_experts_topic" in detail):
+            return {
+                "status": "success",
+                "mode": "onboarding_prompt",
+                "final_response": "TEXT",
+                "reply_text": f"我还没能从这个 `aminer_author_id` 归纳出稳定研究方向，请补充研究方向或代表论文，例如：`/aminer-dp aminer_author_id: {parsed['aminer_author_id']} topics: 多模态, 智能体`。",
+            }
+        if parsed["scholar_name"] and ("profile_unavailable" in detail or "missing_topics" in detail):
+            return {
+                "status": "success",
+                "mode": "onboarding_prompt",
+                "final_response": "TEXT",
+                "reply_text": "我查到了学者线索，但还没成功归纳出稳定研究方向。请补充 `topics`、代表论文标题，或直接描述方向，例如：`/aminer-dp 我是张帆进，清华大学，做多模态智能体和 tool-use`。",
+            }
+        return {
+            "status": "success",
+            "mode": "error",
+            "final_response": "TEXT",
+            "reply_text": f"推荐流程执行失败，出错阶段：{detail}",
+        }
+
+    result: dict[str, Any] = {
+        "status": "success",
+        "mode": pipeline_result.get("mode", "success"),
+        "parsed_input": parsed,
+        "artifacts": {
+            "runtime_config": str(runtime_config_path),
+            "output_dir": str(output_dir),
+        },
+        "pipeline": pipeline_result,
+        "final_response": pipeline_result.get("final_response", "TEXT"),
+    }
+    reply_text = pipeline_result.get("reply_text", "")
+    if reply_text:
+        result["reply_text"] = reply_text
+    return result
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Parse and run aminer-daily-paper recommendation from trigger text.")
+    parser.add_argument("--base-dir", type=Path, default=Path(__file__).resolve().parents[1])
+    parser.add_argument("--text", required=True)
+    parser.add_argument("--config", type=Path, default=None)
+    args = parser.parse_args()
+
+    result = handle_trigger(
+        base_dir=args.base_dir.resolve(),
+        text=args.text,
+        config_path=args.config.resolve() if args.config else None,
+    )
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

skills/aminer-daily-paper/scripts/rec5_api.py

@@ -0,0 +1,210 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import json
+import os
+import ssl
+import time
+import urllib.error
+import urllib.request
+from typing import Any
+
+from scripts.constants import AMINER_PAPER_URL_TEMPLATE, DEFAULT_REC5_URL
+
+DEFAULT_TIMEOUT_SECONDS = 30
+DEFAULT_RETRY_ATTEMPTS = 2
+RETRYABLE_HTTP_CODES = {429, 500, 502, 503, 504}
+
+
+def _clean_text(value: Any) -> str:
+    return " ".join(str(value or "").split()).strip()
+
+
+def resolve_token(config: dict[str, Any] | None = None) -> str:
+    config = config or {}
+    aminer_config = config.get("aminer") if isinstance(config.get("aminer"), dict) else {}
+    return _clean_text(os.getenv("AMINER_API_KEY") or aminer_config.get("token"))
+
+
+def resolve_rec5_url(config: dict[str, Any] | None = None) -> str:
+    config = config or {}
+    aminer_config = config.get("aminer") if isinstance(config.get("aminer"), dict) else {}
+    return _clean_text(aminer_config.get("rec5_url") or os.getenv("AMINER_REC5_URL")) or DEFAULT_REC5_URL
+
+
+def build_api_request(
+    *,
+    aminer_author_id: str = "",
+    author_name: str = "",
+    author_org: str = "",
+    topics: list[str] | None = None,
+    size: int = 5,
+    offset: int = 0,
+    start_year: int | None = None,
+    end_year: int | None = None,
+    language_sort: str = "",
+) -> dict[str, Any]:
+    params: dict[str, Any] = {}
+    if _clean_text(aminer_author_id):
+        params["aminer_author_id"] = _clean_text(aminer_author_id)
+    if _clean_text(author_name):
+        params["author_name"] = _clean_text(author_name)
+    if _clean_text(author_org):
+        params["author_org"] = _clean_text(author_org)
+    cleaned_topics = [_clean_text(t) for t in (topics or []) if _clean_text(t)]
+    if cleaned_topics:
+        params["topics"] = cleaned_topics
+    params["size"] = max(1, min(int(size), 20))
+    params["offset"] = max(0, min(int(offset), 100))
+    if start_year is not None:
+        params["start_year"] = int(start_year)
+    if end_year is not None:
+        params["end_year"] = int(end_year)
+    if _clean_text(language_sort) in {"zh", "en"}:
+        params["language_sort"] = _clean_text(language_sort)
+    return params
+
+
+def normalize_rec5_paper(raw: dict[str, Any]) -> dict[str, Any]:
+    """Normalize raw rec5 paper dict to the in-skill record shape (Markdown display / JSON 输出)."""
+    paper_id = _clean_text(raw.get("paper_id") or raw.get("id"))
+    links = raw.get("links") if isinstance(raw.get("links"), dict) else {}
+    aminer_url = (
+        _clean_text(links.get("aminer"))
+        or _clean_text(raw.get("paper_url"))
+        or (AMINER_PAPER_URL_TEMPLATE.format(paper_id=paper_id) if paper_id else "")
+    )
+    arxiv_url = _clean_text(links.get("arxiv") or raw.get("arxiv_url") or "")
+    pdf_url = _clean_text(links.get("pdf") or raw.get("pdf_url") or "")
+    arxiv_id = _clean_text(raw.get("arxiv_id") or "")
+
+    raw_ss = raw.get("structured_summary")
+    if isinstance(raw_ss, dict):
+        structured_summary: dict[str, str] = {k: _clean_text(v) for k, v in raw_ss.items() if _clean_text(v)}
+    else:
+        structured_summary = {}
+
+    raw_fa = raw.get("famous_authors")
+    famous_authors: list[Any] = []
+    if isinstance(raw_fa, list):
+        for item in raw_fa:
+            if isinstance(item, dict):
+                name = _clean_text(item.get("name"))
+                if not name:
+                    continue
+                famous_authors.append(
+                    {
+                        "name": name,
+                        "description": _clean_text(item.get("description") or item.get("bio") or ""),
+                        "profile_url": _clean_text(item.get("profile_url") or ""),
+                    }
+                )
+            elif isinstance(item, str) and _clean_text(item):
+                famous_authors.append(_clean_text(item))
+
+    raw_profiles = raw.get("aminer_author_profiles")
+    aminer_author_profiles: list[dict[str, Any]] = (
+        [p for p in raw_profiles if isinstance(p, dict)] if isinstance(raw_profiles, list) else []
+    )
+
+    raw_entries = raw.get("author_entries")
+    author_entries: list[dict[str, Any]] = (
+        [e for e in raw_entries if isinstance(e, dict)] if isinstance(raw_entries, list) else []
+    )
+
+    year = raw.get("year")
+    if year is not None:
+        try:
+            year = int(year)
+        except (TypeError, ValueError):
+            year = None
+
+    return {
+        "paper_id": paper_id,
+        "arxiv_id": arxiv_id,
+        "aminer_paper_id": paper_id,
+        "aminer_paper_url": aminer_url,
+        "abs_url": arxiv_url,
+        "pdf_url": pdf_url,
+        "title": _clean_text(raw.get("title")),
+        "year": year,
+        "authors": [_clean_text(a) for a in list(raw.get("authors") or []) if _clean_text(a)],
+        "keywords": [_clean_text(k) for k in list(raw.get("keywords") or []) if _clean_text(k)],
+        "summary": _clean_text(raw.get("summary") or ""),
+        "structured_summary": structured_summary,
+        "famous_authors": famous_authors,
+        "aminer_author_profiles": aminer_author_profiles,
+        "author_entries": author_entries,
+        "source": _clean_text(raw.get("source") or "rec5"),
+        "recommendation_reason": _clean_text(raw.get("recommendation_reason") or ""),
+    }
+
+
+def call_rec5_api(
+    params: dict[str, Any],
+    *,
+    token: str,
+    url: str = DEFAULT_REC5_URL,
+    timeout_seconds: int = DEFAULT_TIMEOUT_SECONDS,
+    retry_attempts: int = DEFAULT_RETRY_ATTEMPTS,
+) -> dict[str, Any]:
+    if not _clean_text(token):
+        raise RuntimeError("missing_aminer_api_key")
+
+    body = json.dumps(params, ensure_ascii=False).encode("utf-8")
+    ssl_context = ssl.create_default_context()
+    opener = urllib.request.build_opener(
+        urllib.request.ProxyHandler({}),
+        urllib.request.HTTPSHandler(context=ssl_context),
+    )
+
+    last_error: Exception | None = None
+    for attempt in range(1, retry_attempts + 2):
+        request = urllib.request.Request(
+            url,
+            data=body,
+            headers={
+                "Content-Type": "application/json;charset=utf-8",
+                "Authorization": token,
+                "User-Agent": "aminer-rec/1.0",
+                "X-Platform": "openclaw",
+            },
+            method="POST",
+        )
+        try:
+            with opener.open(request, timeout=timeout_seconds) as response:  # nosec B310
+                payload = json.loads(response.read().decode("utf-8"))
+        except urllib.error.HTTPError as exc:
+            if exc.code in RETRYABLE_HTTP_CODES and attempt <= retry_attempts:
+                last_error = exc
+                time.sleep(0.5 * attempt)
+                continue
+            raise RuntimeError(f"rec5_api_http_{exc.code}") from exc
+        except Exception as exc:
+            if attempt <= retry_attempts:
+                last_error = exc
+                time.sleep(0.5 * attempt)
+                continue
+            raise RuntimeError(f"rec5_api_error:{exc.__class__.__name__}") from exc
+
+        if not payload.get("success"):
+            msg = _clean_text(payload.get("msg") or str(payload.get("code") or "api_error"))
+            raise RuntimeError(f"rec5_api_failed:{msg}")
+
+        data = payload.get("data")
+        if isinstance(data, list) and data:
+            papers = list(data[0].get("papers") or [])
+            data_obj = data[0] if isinstance(data[0], dict) else {}
+        elif isinstance(data, dict):
+            papers = list(data.get("papers") or [])
+            data_obj = data
+        else:
+            papers = []
+            data_obj = {}
+        analyzed_topics = list(data_obj.get("analyzed_topics") or []) if isinstance(data_obj, dict) else []
+        return {
+            "papers": [p for p in papers if isinstance(p, dict)],
+            "analyzed_topics": [str(t).strip() for t in analyzed_topics if str(t).strip()],
+        }
+
+    raise RuntimeError(f"rec5_api_unreachable:{_clean_text(str(last_error))}") from last_error

skills/aminer-daily-paper/scripts/run_pipeline.py

@@ -0,0 +1,231 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+if __package__ in {None, ""}:
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from scripts.common import write_json
+from scripts.constants import DEFAULT_TOP_K
+from scripts.rec5_api import (
+    build_api_request,
+    call_rec5_api,
+    normalize_rec5_paper,
+    resolve_rec5_url,
+    resolve_token,
+)
+
+
+def _clean_text(value: Any) -> str:
+    return " ".join(str(value or "").split()).strip()
+
+
+def _load_yaml(path: Path | None) -> dict[str, Any]:
+    if path is None or not path.exists():
+        return {}
+    return yaml.safe_load(path.read_text(encoding="utf-8")) or {}
+
+
+def _stage_error(stage: str, detail: Any) -> RuntimeError:
+    compact = _clean_text(str(detail)) or "unknown_error"
+    return RuntimeError(f"{stage}_failed:{compact}")
+
+
+def _format_papers_as_markdown(papers: list[dict[str, Any]], profile_topics: list[str]) -> str:
+    """Render recommended papers as Markdown for the host to display."""
+    lines: list[str] = []
+    topic_hint = " / ".join(profile_topics[:5]) if profile_topics else ""
+    header = f"为你推荐 {len(papers)} 篇相关论文"
+    if topic_hint:
+        header += f"（研究方向：{topic_hint}）"
+    lines.append(header)
+
+    for idx, paper in enumerate(papers, start=1):
+        lines.append("")
+        lines.append("---")
+        lines.append("")
+        title = _clean_text(paper.get("title") or "")
+        url = _clean_text(paper.get("aminer_paper_url") or paper.get("abs_url") or "")
+        title_line = f"**{idx}. [{title}]({url})**" if url else f"**{idx}. {title}**"
+        lines.append(title_line)
+
+        year = paper.get("year")
+        keywords = paper.get("keywords") or []
+        authors = paper.get("authors") or []
+        summary = _clean_text(paper.get("summary") or "")
+
+        meta_parts: list[str] = []
+        if year:
+            meta_parts.append(f"年份：{year}")
+        if keywords:
+            meta_parts.append(f"关键词：{' / '.join(str(k) for k in keywords[:5])}")
+        if meta_parts:
+            lines.append(" | ".join(meta_parts))
+
+        if authors:
+            author_str = "、".join(str(a) for a in authors[:6])
+            if len(authors) > 6:
+                author_str += " et al."
+            lines.append(f"作者：{author_str}")
+
+        if summary:
+            truncated = summary if len(summary) <= 300 else summary[:300].rstrip() + "…"
+            lines.append("")
+            lines.append(truncated)
+
+    return "\n".join(lines)
+
+
+def _topics_from_paper_titles(paper_titles: list[str]) -> list[str]:
+    """Extract simple topic hints from paper titles when no explicit topics provided."""
+    topics: list[str] = []
+    seen: set[str] = set()
+    for title in paper_titles:
+        cleaned = _clean_text(title)
+        if not cleaned:
+            continue
+        candidate = cleaned.split(":")[0].strip() if ":" in cleaned else cleaned
+        if candidate and candidate.casefold() not in seen and len(candidate) <= 80:
+            seen.add(candidate.casefold())
+            topics.append(candidate)
+    return topics[:5]
+
+
+def run_pipeline(
+    *,
+    output_dir: Path,
+    config: dict[str, Any],
+    aminer_author_id: str,
+    topics: list[str],
+    scholar_name: str,
+    scholar_org: str,
+    paper_titles: list[str],
+    papers_file: str,
+    free_text: str,
+    language_sort: str = "",
+    size: int = 0,
+) -> dict[str, Any]:
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    token = resolve_token(config)
+    if not token:
+        raise _stage_error("auth", "AMINER_API_KEY is not set")
+
+    all_topics = list(topics)
+    if paper_titles and not all_topics and not scholar_name and not aminer_author_id:
+        all_topics = _topics_from_paper_titles(paper_titles)
+
+    search_config = config.get("search") if isinstance(config.get("search"), dict) else {}
+    if size <= 0:
+        size = max(1, min(int(search_config.get("top_k") or DEFAULT_TOP_K), 20))
+
+    # language_sort only when user explicitly passed language_sort: zh|en in the trigger (see SKILL).
+    sort_for_api = _clean_text(language_sort) if _clean_text(language_sort) in {"zh", "en"} else ""
+
+    api_request = build_api_request(
+        aminer_author_id=aminer_author_id,
+        author_name=scholar_name,
+        author_org=scholar_org,
+        topics=all_topics,
+        size=size,
+        language_sort=sort_for_api,
+    )
+
+    try:
+        result = call_rec5_api(api_request, token=token, url=resolve_rec5_url(config))
+        raw_papers = result["papers"]
+        api_analyzed_topics = result.get("analyzed_topics") or []
+    except Exception as exc:
+        raise _stage_error("recall", exc) from exc
+
+    if not raw_papers:
+        raise _stage_error("recall", "no_papers_returned")
+
+    papers = [normalize_rec5_paper(p) for p in raw_papers]
+    papers = [p for p in papers if _clean_text(p.get("title"))]
+
+    profile_topics = api_analyzed_topics or all_topics or ([scholar_name] if scholar_name else [])
+    summarized_payload = {
+        "status": "success",
+        "profile_topics": profile_topics,
+        "profile_name": scholar_name or "",
+        "profile_source": "scholar_path" if (aminer_author_id or scholar_name) else "topic_path",
+        "papers": papers,
+    }
+
+    mode = "scholar_path" if (aminer_author_id or scholar_name) else "topic_path"
+
+    write_json(
+        output_dir / "request_context.json",
+        {
+            "aminer_author_id": aminer_author_id,
+            "input_topics": topics,
+            "topics": all_topics,
+            "scholar_name": scholar_name,
+            "scholar_org": scholar_org,
+            "paper_titles": paper_titles,
+            "papers_file": papers_file,
+            "free_text": free_text,
+            "language_sort": language_sort,
+            "api_request": api_request,
+        },
+    )
+
+    summarized_path = output_dir / "papers_summarized.json"
+    write_json(summarized_path, summarized_payload)
+    markdown_text = _format_papers_as_markdown(papers, profile_topics)
+    return {
+        "status": "success",
+        "summarized_path": str(summarized_path),
+        "final_response": "TEXT",
+        "reply_text": markdown_text,
+        "mode": mode,
+        "paper_count": len(papers),
+    }
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Run aminer-daily-paper pipeline via rec5 API.")
+    parser.add_argument("--base-dir", type=Path, default=Path(__file__).resolve().parents[1])
+    parser.add_argument("--config", type=Path, default=None)
+    parser.add_argument("--output-dir", type=Path, default=Path(__file__).resolve().parents[1] / "outputs")
+    parser.add_argument("--aminer-author-id", default="")
+    parser.add_argument("--topics", nargs="*", default=[])
+    parser.add_argument("--scholar-name", default="")
+    parser.add_argument("--scholar-org", default="")
+    parser.add_argument("--paper-title", action="append", dest="paper_titles", default=[])
+    parser.add_argument("--papers-file", default="")
+    parser.add_argument("--free-text", default="")
+    parser.add_argument("--language-sort", default="")
+    parser.add_argument("--size", type=int, default=0)
+    args = parser.parse_args()
+
+    resolved_config = args.config.resolve() if args.config else None
+    config = _load_yaml(resolved_config)
+
+    result = run_pipeline(
+        output_dir=args.output_dir.resolve(),
+        config=config,
+        aminer_author_id=args.aminer_author_id,
+        topics=list(args.topics or []),
+        scholar_name=args.scholar_name,
+        scholar_org=args.scholar_org,
+        paper_titles=list(args.paper_titles or []),
+        papers_file=args.papers_file,
+        free_text=args.free_text,
+        language_sort=args.language_sort,
+        size=args.size,
+    )
+    print(json.dumps(result, ensure_ascii=False, indent=2))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())