Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add KCNyu/clawock --skill daily-deep-briefgit clone --depth 1 https://github.com/KCNyu/clawockWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kcnyu/clawock/daily-deep-brief)<a href="https://agentmods.dev/skills/kcnyu/clawock/daily-deep-brief"><img src="https://agentmods.dev/badge/skills/kcnyu/clawock/daily-deep-brief/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kcnyu/clawock/daily-deep-brief"><img src="https://agentmods.dev/badge/skills/kcnyu/clawock/daily-deep-brief.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00163 | $0.23276 |
| Opus 5 | $0.00081 | $0.11638 |
| Sonnet 5 | $0.00033 | $0.04655 |
| Haiku 4.5 | $0.00016 | $0.02328 |
Grade A, and why
daily-deep-brief scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 1,011 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Daily Deep Brief (08:00 HKT, weekday)
8 点这个时点:HK 开盘前 ~90 分钟,US 已收盘 ~4 小时。盘前是 deep think 最好的窗口 — 有完整夜间消息面,没有盘中执行压力。
Harness 架构(不可越权)
┌──────────────────────┐ ┌────────────────┐ ┌──────────────────────────┐
│ clawock brief preflight │──►│ LLM (你 / Rick) │──►│ clawock brief postflight │
│ (确定性 + 幂等) │ │ (只写判断/反方) │ │ 校验 → 渲染 → 投递 │
└──────────────────────┘ └────────────────┘ └──────────────────────────┘
指标/因子/风险/evidence 查 packet summary 校验 plan+judgment,
编译 typed packet 写 plan + judgment brief_render 出 md+微信卡,
(纯文本,不写版式) Pages 只读稳定 projection
为什么这样分:之前完全交给 LLM,模型可能漏快照、漏 HHI、漏 FX、漏 retrospective。 确定性步骤交给脚本(一定执行,无遗漏);LLM 只做"分析综合"这个不能脚本化的部分。
🔒 Exec 铁律:退出码就是整回合的判据
每一条 exec 的退出码会被当成整个回合的成败。 一条中途的非零退出会让 cron 把
已经做完并已投递的一轮判成 error(真实案例:2026-08-18,简报全部产出、
wechat_sent: true、commit+push 都成了,运行时 trace.artifacts.finalStatus 也是
success,但 08:14 那条 exec 非零退出,整轮仍被记成红)。
2>/dev/null 只吞 stderr,不改退出码。 这是最容易踩的一条:
# ❌ 踩过的原文(2026-08-18):末尾那个文件当天不存在 → head exit 1 → 整回合判红
ls -la memory/2026-08-17* 2>/dev/null; echo "---"; \
head -c 2500 memory/2026-08-17-pre-open.md 2>/dev/null; echo "==="; \
head -c 1500 memory/2026-08-17.md 2>/dev/null
# ✅ 读"可能不存在"的文件时,自己兜底退出码
ls -la memory/2026-08-17* 2>/dev/null || true; echo "---"; \
head -c 2500 memory/2026-08-17-pre-open.md 2>/dev/null || true; echo "==="; \
head -c 1500 memory/2026-08-17.md 2>/dev/null || true
三条硬规则:
- 凡是读可选文件/通配符(日报、昨日 notes、
.tmp/产物)都要|| true。 日期型文件天生可能缺(当天没写 notes、休市、上游没产出),缺失不是错误。 ;串起来的复合命令,最后一条决定退出码 —— 前面成不成功都不算。 拿不准就在整条命令末尾补; true。grep/test/head/ls用作"看一眼"而不是"判定"时,一律兜底。grep -q找不到东西返回 1 是正常语义,不该让一轮简报变红。
配套的窄规则见 Step 5 的「送达确认铁律」(#558):那条禁的是 postflight 之后 再去读 marker 确认送达;本节是它的一般形式 —— 别让"看一眼"的退出码决定一轮的成败。
6-step 流程(严格按顺序)
Step 1: 跑 preflight(一行命令搞定所有确定性活)
clawock brief preflight
这一步内部做了:
clawock analyze-us— US 价格刷新(7-route fallback + RSI/MA)clawock analyze-hk— HK 价格刷新(Tencent 主源 + Eastmoney 全量独立对账/兜底 → stooq → yfinance + 恒指 + 信号)clawock fx --json— USDHKD 实时汇率(Frankfurter → exchangerate.host → Yahoo)cp portfolio.json memory/snapshots/{date}.json— 每日快照(longitudinal 基础设施)- 为什么:
portfolio.json是滚动覆盖的 ground truth,每次刷价就丢前一刻状态。 有 snapshot 历史才能做 Rolling P&L 曲线 / Alpha vs benchmark / Drawdown 分析 / Position 变化追溯。 - ⚠️ 不可补做 — 每过一天少一份永远拿不回来的数据。
- 为什么:
- HHI / Top2 集中度算法(HK + US leg 分开)
- SEC EDGAR fundamentals — 从当天
portfolio.json动态筛选shares > 0的 US 持仓;跳过is_leveraged_etf=true或被名称启发式识别为杠杆 ETF 的标的,其余单股逐一跑clawock filings- 杠杆 ETF 检测启发式(name 关键词):'倍', 'Direxion', 'T-Rex', 'Defiance', 'ProShares', '2X Long', '3X Long', 'Daily Target'
- 不在文档硬编码 ticker;实际名单以当天持仓和上述动态过滤结果为准
- Retrospective:读 prior v2 plan,对每个 decision 按 strategy 检查 condition 是否触发、模拟 benefit 与 confidence calibration
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday Changed · -1 lines 76fcd478f14c
- 2d ago Changed · +3 lines b2ab789893da
- 6d ago Changed · +23 lines · +19 tokens per session cdcbc8fb4122
- 10d ago First seen · 986 lines · 144 tokens per session scan A d6ff8c8b9ecf
daily-deep-brief is a skill published in the GitHub repository KCNyu/clawock (14 stars, last pushed today), licensed MIT. It adds 163 tokens to every session and 23,276 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
tushare
A Python interface to Tushare, a financial-data service providing market, company, fund, futures, and other finance-related data. It shows how to authenticate and query the service into pandas tables.
financial-expense-automation
An expense-receipt workflow that reads PDF and image attachments, extracts their details, checks whether they are valid expense receipts, and records them in Feishu Bitable, a cloud table service.
credit-analysis
A framework for analysing bonds and other fixed-income investments, including issuer credit quality, interest-rate sensitivity, credit spreads, and default risk. A credit rating estimates how likely an issuer or specific bond is to repay its debt.
crypto-monitor
A scheduled cryptocurrency price checker that uses CoinGecko, a service providing market data. It tracks Bitcoin, Ethereum, and Litecoin in US dollars and Chinese yuan, including their change over the past 24 hours.
vibe-trading
Professional finance research toolkit — backtesting (7 engines + benchmark comparison panel), factor analysis, Alpha Zoo (452 pre-built alphas across qlib158/alpha101/gtja191/academic), options pricing, 79 finance skills, 29 multi-agent swarm teams, Trade Journal analyzer, and Shadow Account (extract → backtest →…
quant-statistics
Quantitative statistical methods: ADF unit-root / cointegration tests, GARCH volatility modeling, regression diagnostics (heteroskedasticity / autocorrelation), Bootstrap, and hypothesis testing.