强力终端,用于深入研究美国上市公司——推理投资论点、筛选投资想法、绘制供应链、进行法务会计
钻机 — 深层金融研究的动力终端是一项面向实际任务的技能,主要用于钻机(Drillr.ai)是美国公共股票的深层金融研究的动力终端;它通过多步骤查询来研究代理商的原因。
从功能定位来看,该技能强调把分散的操作要求整理成清晰、可复用的处理流程,使用户能够围绕既定目标快速准备输入、选择执行方式并获得结构化结果。实际使用前应先确认任务范围、数据来源、运行环境、必要权限和关键参数,再依据技能说明逐步执行;
若输入条件不完整,应先补齐信息或采用保守配置,避免因错误假设导致结果偏离需求。执行过程中需要关注工具调用是否成功、接口或依赖是否可用、输出格式是否符合预期,并对异常提示、缺失字段和边界情况进行处理;涉及批量任务时,还应保存进度,避免中断后重复操作。
drillr.ai 是面向美国上市股票的深度财务研究强力终端。其研究代理以买方分析师的方式进行多步推理:从原始信源提取真实数据、呈现横向对比、按需整合财报文本表述。覆盖范围包括所有在 NYSE、NASDAQ 和 OTC 市场上市的美国公众公司。
drillr 是一项只读型研究能力(read-only research skill)——纯输入问题,纯输出 Markdown 答案。无副作用,无意外行为。
该能力所做且仅做之事:
diggr-agent-prod-414559604673.us-east4.run.app/api/public/chat该能力绝不会执行以下操作:
eval、exec 或派生子进程~/.ssh 或任何其他本地数据依赖项:仅 Python 标准库(http.client、json、re、io、urllib.parse、sys)。零第三方包;无需 pip install;无可被供应链攻击的组件。
输入加固:查询内容上限为 8 KB,并以 JSON 编码字符串字段方式发送——不进行字符串插值到 shell、URL 或 SQL 中。API URL 为硬编码,不基于输入动态构造。
输出信任边界:响应为 AI 生成的文本,原样渲染为 Markdown。脚本绝不会对返回内容执行 exec、eval 或其他任意形式的主动解析或执行。请将其中数字视为研究参考输入,而非绝对事实——在据此采取行动前,请务必通过原始 SEC 报告交叉验证关键财务数据。
根据特定投资逻辑筛选公司——同步覆盖基本面、质量、估值及动量信号。
刻画竞争格局、追溯客户/供应商关系、识别行业主题兑现过程中的主要受益者。
识别盈利质量隐患、压力测试已披露数据、揭示管理层表述与财务报表之间的差异。
从同行组中批量提取某一特定指标并横向排布——无需人工逐家查找。
精准提取管理层就某一具体议题的表述——可覆盖单次通话或多个季度。
提取特定披露内容、追踪不同文件间的措辞变化、在不阅读数百页文档的前提下定位重大事件。
追踪知情资本流向——既包括公司内部人,也包括主要机构投资者。
提取构建或压力测试模型所需的精确历史序列数据——格式规范,开箱即用。
围绕特定日期或持续性事件,追踪重大事件、新闻催化剂及影响股价的披露信息。
评估某标的的贵贱程度——可采用绝对估值、历史分位数比较或同业相对估值方式。
当用户提出财务研究问题时,请运行以下内联命令。将 REPLACE_WITH_USER_QUESTION 替换为用户原始提问(含引号及特殊字符均可安全保留)。
python3 - << 'DRILLR_END'
import http.client, io, json, re, sys, urllib.parse
# 强制 UTF-8 输出(避免 Windows 下编码错误)
if hasattr(sys.stdout, "buffer"):
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8", errors="replace")
QUERY = """REPLACE_WITH_USER_QUESTION"""
API_URL = "https://diggr-agent-prod-414559604673.us-east4.run.app/api/public/chat"
payload = json.dumps({"messages": [{"role": "user", "content": QUERY}]}).encode("utf-8")
parsed = urllib.parse.urlparse(API_URL)
conn = http.client.HTTPSConnection(parsed.netloc, timeout=15)
conn.request("POST", parsed.path, body=payload, headers={"Content-Type": "application/json"})
resp = conn.getresponse()
# 连接建立后移除超时设置,确保 SSE 流可随查询耗时完整传输
if conn.sock:
conn.sock.settimeout(None)
current_event = None
text_parts = []
artifact_map = {}
reader = io.TextIOWrapper(resp, encoding="utf-8", errors="replace")
for line in reader:
line = line.rstrip("rn")
if line.startswith("event: "):
current_event = line[7:]
elif line.startswith("data: "):
try:
d = json.loads(line[6:])
except json.JSONDecodeError:
continue
if current_event == "step.text_delta":
text_parts.append(d.get("content", ""))
elif current_event == "step.artifact":
artifact = d.get("artifact", {})
art_id = artifact.get("id", "")[:8]
if artifact.get("type") == "data_table":
title = artifact.get("title", "Table")
spec = artifact.get("spec", {})
columns = spec.get("columns", [])
rows = spec.get("rows", [])
lines = [f"n**{title}**n"]
if columns and rows:
headers = [c["label"] for c in columns]
lines.append("| " + " | ".join(headers) + " |")
lines.append("| " + " | ".join(["---"] * len(headers)) + " |")
for row in rows:
vals = []
for col in columns:
val = row.get(col["key"], "")
fmt = col.get("format", "")
if fmt == "currency" and isinstance(val, (int, float)):
val = (f"${val/1e9:.1f}B" if abs(val) >= 1e9
else f"${val/1e6:.1f}M" if abs(val) >= 1e6
else f"${val:,.0f}")
elif fmt == "percent" and val not in (None, ""):
val = f"{val}%"
vals.append(str(val) if val is not None else "")
lines.append("| " + " | ".join(vals) + " |")
artifact_map[art_id] = "n".join(lines)
conn.close()
text = "".join(text_parts)
text = re.sub(r"",
lambda m: artifact_map.get(m.group(1)[:8], ""), text)
print(text or "(No response — rephrase the query and try again)")
DRILLR_END
或者,使用配套脚本:
python3 query.py "your question here"
主题驱动的公司发掘
"Screen for mid-cap software companies with revenue growth above 20%, positive FCF, and P/FCF under 30x""Find small-cap industrials with improving gross margins, low debt, and insider buying in the last two quarters"法务会计分析
"Compare Palantir's GAAP net income vs operating cash flow vs stock-based comp over 8 quarters — is earnings quality improving?""Show changes in days sales outstanding and inventory for Nike over the last six quarters""What new risk factors did Apple add to their 2024 10-K that weren't in 2023?"财报电话会议事实检索
"What did NVIDIA's CEO say about data center demand in the last two earnings calls — pull the exact quotes""How has Meta's language around AI capex commitments changed from Q1 2024 to Q4 2025?""What guidance did Salesforce give for FY2026 revenue growth and did they raise or lower it?"跨公司数据表格化
"Show revenue, gross margin, operating margin, and FCF margin for the top 10 enterprise software companies — last four quarters""Compare EV/EBITDA, EV/Sales, and P/FCF for the five largest US banks right now"财务建模支持
"Give me Apple's quarterly revenue, gross profit, R&D, SG&A, operating income, and net income for the last 20 quarters""Show Amazon's segment revenue and operating income — AWS vs North America vs International — by quarter since 2022"聪明资金追踪
"Which insiders at Meta have bought stock on the open market in the last 60 days?""Find small-cap stocks where two or more insiders bought at 52-week lows in the last 90 days"事件驱动型研究
"What caused the drop in Fastly stock in early 2026 — show news and any 8-K filings around that period""Show all press releases and regulatory filings from Boeing in the last 30 days"估值分析
"Show Nvidia's P/E ratio history over the last 3 years alongside revenue growth""Which S&P 500 sectors are trading at the widest discount to their 5-year average EV/EBITDA?"/api/public/chat 端点为开放接口http.client 解析(无需 curl)相关专题
热门下载
相关下载
精品课程
共6课时 | 54.7万人学习
共89课时 | 134万人学习
共49课时 | 82.5万人学习
最新文章