← ClaudeAtlas

investment-scorecardlisted

Convert AI product judgment and fact packs into the library's investment scorecard layer. Use for AI / OSS / infrastructure investability scoring, macro gate, deal screening, one-vote veto checks, pass / watch / recommend decisions, and scorecard handoff into IC memo.
qihangzhang-272/agent-skill-library · ★ 1 · DevOps & Infrastructure · score 60
Install: claude install-skill qihangzhang-272/agent-skill-library
# Investment Scorecard This node replaces direct use of `oss-investment-scorecard` in the library's default workflow. Use it after `investment-ai-product-judgment` and before DD / IC memo writing. ## References Always read: - `references/oss-investment-scorecard.md` Read when needed: - `references/scored-examples.md` for calibration. - `references/deal-screening.md` for pass / further diligence / hard pass framing. - `references/cases/hugging-face.md` and `references/cases/inferact-vllm.md` for example structure. - `references/template/evaluation-template.md` for structured scorecard output. ## Output ```text Scorecard reference usage: Macro gate: Deal-screening verdict: Dimension scores: Weighted total: One-vote veto check: Investment recommendation: IC thesis seed: DD priorities: Watch triggers: Handoff to IC memo: ``` ## 完成标准 - 声明 as-of 日期、币种和单位、投资问题、阶段、已提供的基金标准,以及 OSS 或非 OSS 的五维框架适配 - 逐项完成 macro gate;Q1/Q2 的 NO 与 Q3 的风险语义不得混淆,并说明是否只保留诊断性评分 - 对生态或采用、团队与全球化、技术护城河与定位、商业化与 PMF、资本退出路径五维逐项连接证据和反证 - 权重、原始分、加权分和总分计算可复核,缺失数据的处理不制造虚假精度 - 只使用明示的一票否决项,并把否决、重大风险、总分和基金标准显式协调 - 形成一致的建议、DD 优先级和可测量跟踪触发器,并列出会改变结论的证据 - 经合理查找仍不可得的信息必须用业务语言写清缺失事项、不可得原因、对结论的影响、当前保守处理和重新核验条件后继续;缺失数据不得被写成零或虚假精确分数 - 形成可独立检查的产物:$artifact。 - 在 Mode 内调用与独立调用执行相同标准;无法满足的项目必须说明缺口与影响,不得伪造完成。