📦 Banker — 银行家

v0.9.6

Memo 从 CN raw-data/ 快照生成投行级研究备忘录（analysis.md + slides-outline.md + data-provenance.md）。在用户提出以下请求时使用……

0· 19·0 当前·0 累计

by @jackdark425 (jackdark)

开发工具代码生成数据分析数据可视化网络工具

下载技能包

最后更新

2026/4/20

安全扫描

VirusTotal

无害

查看报告

OpenClaw

可疑

medium confidence

The skill's stated goal (produce provenance-backed banker memos from raw JSON snapshots) is plausible, but the instructions and files are inconsistent about how the agent actually receives the raw data and about required tooling — this mismatch could cause the agent to hallucinate numbers or require undeclared binaries/scripts.

评估建议

This skill intends to produce provenance-backed banker memos, but the provided prompt-builder only lists filenames rather than embedding raw JSON content. Before installing/using it, verify the following: (1) Confirm how the agent will access the actual raw-data JSON (is the openclaw agent invocation able to read {raw_dir} on disk, or must you embed file contents in the prompt?). If the agent cannot access the files, it may fabricate numbers and provenance. (2) Ensure the runtime environment sup...

详细分析 ▾

⚠ 用途与能力

The skill claims to produce numerically-traceable banker memos from raw-data/*.json snapshots. That purpose legitimately needs access to the raw JSON contents and the validation/build tools referenced. However the declared requirements list no binaries, env vars, or config paths, yet the README and runtime steps expect external tools (openclaw agent CLI, validate-delivery.py, build_deck.py and the cn-client-investigation pipeline). The lack of declared dependencies is disproportionate to the stated purpose.

⚠ 指令范围

SKILL.md and the prompt template repeatedly state the agent must base every hard number on the raw-data files. But the provided helper script only enumerates filenames (file_list) and does not embed or pass raw JSON contents into the prompt; the dispatch step runs `openclaw agent --message "$(cat prompt.md)"`, which would only supply the prompt text. Unless the runtime agent/CLI has filesystem access to {raw_dir} (not declared), the agent will not actually see the JSON and may invent numeric provenance. The instructions also reference external validation and build scripts (validate-delivery.py, build_deck.py) without packaging or declaring them.

ℹ 安装机制

The skill is instruction-only with one helper script — no install spec is present, which lowers risk. However because it requires running external scripts/CLI tools at runtime (openclaw agent, validate-delivery.py, build_deck.py) there is an undeclared installation/runtime dependency surface that deployers must satisfy.

✓ 凭证需求

The skill declares no environment variables, credentials, or config paths. That is appropriate for its stated purpose. The script reads local filesystem paths (raw-data/*.json) which is expected. There is no request for unrelated secrets or remote tokens in the manifest.

✓ 持久化与权限

always:false and no requests to modify other skills or global agent configuration. The skill does invoke autonomous agent workflows by design (normal for skills), but it does not request elevated persistence privileges.

安全有层次，运行前请审查代码。

运行时依赖

无特殊依赖

版本

latestv0.9.62026/4/20

banker-memo v0.9.6 - Switched from fixed Python templates to prompt-driven memo generation using agent reasoning, enabling richer, peer-benchmarked analysis. - Outputs now include analysis.md (8-section memo), slides-outline.md (PPT blueprint, 10–15 slides), and data-provenance.md (trace every number to raw source). - Hard constraints: every figure traced to raw-data, peer comps tagged as estimates, valuation covers multiple scenarios, and all outputs validated by strict audit gates. - Slide deck outlines are now flexible (not fixed at 8 slides), consumed by build_deck.py v0.9.6+ for custom layout. - Strict requirements on provenance, peer benchmarking, and 4C's credit view; not suitable for US-listed firms or quick fact-sheets.

● 无害

安装命令

点击复制

官方npx clawhub@latest install banker-memo

镜像加速npx clawhub@latest install banker-memo --registry https://cn.longxiaskill.com镜像同步中

需要定制？告诉我你的需求 →

数据来源：ClawHub ↗ · 中文优化：龙虾技能库