agentsclimarketplace

Ir deep review

Skill LechuanWANG/ir-skill/skills/ir-deep-review

Codex skill for local investment research

Install
npx -y skills add LechuanWANG/ir-skill --skill ir-deep-review

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Independent investment research cross-examination for explicitly requested deep research, review, or challenge of evidence, assumptions, valuation, competition, governance, and tail risks.

SKILL.md

3.0 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

深度研究与独立质询

只在用户明确要求深度研究、独立审阅或交叉质询时使用。先读取 ../shared/research-discipline.md,再按持有期调用 ../ir-long-term-trading/SKILL.md../ir-medium-term-catalyst/SKILL.md../ir-short-term-trading/SKILL.md。普通研究不得自动升级为多审阅者流程。

1. 建立共同证据包

主 Agent 先准备研究对象、决策、持有期、as_of、候选和约束、核心假设、原始证据、关键缺口,以及可能改变排序或行动标签的变量。证据包须覆盖已知争议和反方路径;所有审阅者使用同一时间边界,不得引入决策时点之后的信息证明此前判断。

2. 独立审阅

工具和并发允许时,最多启动三位互不查看彼此结论的审阅者,主 Agent 保留裁决职责:

  1. 商业、财务与估值:商业模式、需求、利润/现金质量、资本回报和价格隐含预期。
  2. 竞争与替代解释:竞争者反应、供需周期、扩产降价、技术替代和乐观因果链。
  3. 治理与尾部风险:会计、杠杆、流动性、治理、监管、诚信和永久性损失。

并发不足时合并角色;完全不能独立审阅时,主 Agent 做明确标注的对立案例检查,不得伪称多 Agent 复核。每位审阅者提交当前立场、两至四项关键判断、最强正反证据、最可能遗漏变量、证伪条件和置信度;候选比较同时给出排序或行动倾向。

3. 交叉质询

只转交可能改变核心假设、估值、排序、主要风险或行动标签的冲突。质询应指出具体主张、未证实假设、证据/时间口径问题及其决策影响;原审阅者以补充证据、量化影响、收窄范围、降低置信度、修改排序或撤回主张回应,并标记为 支持收窄未解决撤回

当决策相关分歧已充分检验、没有未处理的高影响证据缺口,且下一轮不再可能改变决策时停止;不为展示讨论延长流程,也不因赶时间保留未检验的关键分歧。

4. 主 Agent 裁决

不按票数或平均分决定,而是综合证据强度、因果关系、影响量级、赔率和错误损失,明确独立证据支持的共识、仍会改变决策的分歧及当前证据更支持的一方、未覆盖盲区和已撤回观点。

输出当前排序或单股判断、主要行动标签、现金/替代项比较、两至四个关键变量、最可能推翻结论的证据、触发条件和复核时间。最终只展示真正改变判断、未解决或界定结论边界的质询与修订,不输出完整对话或模型私有推理过程。

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.