Content & Media Creation · daymade/claude-code-skills
benchmark-due-diligence
Runs adversarial due-diligence on a benchmark the user envies — a founder, KOL, company, or product whose claimed success looks inflated — splitting marketing bubble from real signal, then mapping the validated playbook onto the user's own resources. Use whenever the user wants to 尽调/对标/拆解 a competitor or role-model, 抄/偷师 someone's playbook, suspects 水分/泡沫 in their claims (#1 on Product Hunt, 0-to-1M users, funding, 估值几个亿), asks whether wins are 真本事 vs 运气/时机, or says someone is 太成功了/crushing it and wants the real story — even if they never say 尽调. Prefer over deep-research for debunking inflated claims and extracting a replicable playbook rather than a neutral briefing.
L3 Networked · Declare domain & key scopeV1 Commit-PinnedNetwork AccessDual-pass Audit
Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.
1How It Works & Mechanism
Pure prompt-orchestrated skill: Zero scripts or code in the directory. Its functionality is driven entirely by SKILL.md instructions prompting the model to spawn sub-agents (Workflow/Task tools) and combine other skills/CLIs; it is not an executable program.
daymade-financial/benchmark-due-diligence/SKILL.md
This skill is an **orchestrator** — it spawns parallel collection + verification agents (via the `Workflow` tool, or `Task` agents) and may invoke other skills (`deep-research`, `osint-investigate`, `qcc`).
What is happening here: Explains 'why the capability works'—functionality stems from orchestration discipline and prompt persona/stance design rather than code. The same section also stipulates that sub-agents cannot spawn further sub-agents or invoke skills, strictly prohibiting `context: fork` (which would silently break the fan-out).
The core mechanism is adversarial bubble-busting: It defaults to assuming promotional figures are inflated until proven true by evidence, and then maps surviving validated insights into actionable steps for the commissioner—acting as the flip side to neutral deep-research.
daymade-financial/benchmark-due-diligence/SKILL.md
this skill **assumes the picture is inflated until proven otherwise** and converts the survivors into the commissioner's own moves.
'Assuming inflation until falsified' is the fundamental distinction from standard due diligence/research skills. It answers three questions: how much is real vs. marketing hype, how much is replicable vs. luck/timing, and what specific actions the commissioner can take with it.
Four-stage execution architecture: Dimension-by-dimension 'collection → adversarial verification' pipeline (no global barrier; verified immediately upon collection) → single synthesis agent generating DD conclusions → dedicated mapping agent deriving 'what this means for you'.
daymade-financial/benchmark-due-diligence/references/workflow_orchestration_template.md
log('Phase 1+2: fan-out collect → adversarial verify (bubble-bust)')
The reference document provides a ready-to-fill Workflow script (independent agents and schemas for collection and verification, across 6 dimensions with ~14 agents). SKILL.md specifies that 'collectors maintain an objective stance, verifiers maintain default skepticism, tagging findings item-by-item as L1–L4 and ruling them Confirmed / Largely Credible / Questionable / Debunked (Inflated)'.
Privacy core relies on dual-channel injection isolation: Public FACTS are injected into all agents, while the confidential COMMISSIONER_CONTEXT is fed solely into the final mapping agent—preventing commissioner client names or strategies from leaking into public WebSearch queries.
daymade-financial/benchmark-due-diligence/SKILL.md
If the commissioner's client names or strategy leak into those prompts, they get searched on the open web — a privacy breach.
The 'channel split' is designated as the primary rule for protecting the commissioner. Phase 0 also requires verifying the entity relationship graph and headline claim attribution upfront (confirming whether the trophy stat was genuinely achieved by the subject), strictly forbidding ownership inferences based on domains or names.
Evidence grading discipline is built into the pipeline: Every finding must include a source URL + source_kind (self-reported/marketing vs. independent third-party vs. mixed); filling gaps with speculation is prohibited. The verification stage actively searches for disconfirming evidence.
daymade-financial/benchmark-due-diligence/references/evidence_grading_rubric.md
Bubble-busting is the act of moving a claim *down* from its self-asserted level.
L4 = verifiable hard data / L3 = multiple independent sources / L2 = single credible third party / L1 = self-reported marketing only. Claims like '#1 of the year' are typically downgraded by official L4 platform records to 'category award + monthly #1'.
2Core Capabilities
01Adversarially breaks down envied benchmarks (founders/KOLs/companies/products), distinguishing real substance from marketing hype
02Headline claim attribution verification (#1 to-verify): Identifies who actually drove '0→1M users' and what the true scale was
03Multi-dimensional parallel collection + just-in-time adversarial verification pipeline (L1–L4 grading and rulings)
04Attribution deconstruction: Calculates the breakdown of success across product, timing, personal branding, and operations (percentage ranges + rationale), separating replicable methods from luck or innate endowment
05Playbook mapping to commissioner's actual resources: 4-tag resource mapping matrix (Adoptable / Irreplicable / Already Doing / Bubble-Do Not Copy) + actionable implementation checklist
06Bubble-busting verification checklist: Claim | Evidence Grade | Ruling | Basis, sorted in descending order of inflation
3External Dependencies
| Kind | Dependency / Endpoint |
|---|---|
| cli | agent-reach |
| api | WebSearch / WebFetch（宿主工具，由派生的采集/验证 agent 执行） |
| package | deep-research / osint-investigate / qcc 技能族（编排内组合调用，非本目录自带） |
| api | Workflow / Task agent 工具（宿主扇出原语） |
4Recommended Governance & Evidence
Suggested Controls for Safe Operation
For Agent Runtime
Declare allowed domains and key scope
For Administrators
Least privilege keys, retain request logs
Evidence & Triggers (File Paths & Excerpts)
- 网络端点: Declares external network access: Spawns agents that perform external retrieval via WebSearch/WebFetch (queries depend on the user-specified research target and destinations cannot be pre-enumerated).
Audit Notes: Consistent with the Yellow rating of deep-research in the same repository: Pure prompt orchestration, zero scripts, zero credential access, and zero local disk writes. However, web research is a core capability, outbound targets depend on the research subject and cannot be pre-enumerated, and scraped content flowing back into the model context presents a prompt injection attack surface. Additionally, there is a privacy consideration where commissioner private context could leak into external WebSearch queries if mistakenly passed to collection/verification agents (mitigated by built-in dual-channel isolation, but still a soft constraint relying on model compliance).
5Second-Pass Verification
- [ok] 『编排器 + 禁 context: fork』的运行约束 — SKILL.md 原文 'Subagents cannot spawn subagents or call skills. Setting `context: fork` would silently break the entire fan-out.' 存在，与 deep-research/osint-investigate 同约束叙事一致。
- [ok] 双通道 FACTS/COMMISSIONER_CONTEXT 注入隔离 — SKILL.md 通道表与 workflow 模板注释（'PRIVATE commissioner reality — injected into the Phase-4 mapping agent ONLY'）逐字一致；Trap 6 在 evidence_discipline_traps.md 有专节。
- [ok] L1–L4 分级与裁决词存在于 rubric — evidence_grading_rubric.md 含 L4/L3/L2/L1 定义表与 坐实/大体可信/存疑/证伪 裁决；'Bubble-busting is the act of moving a claim *down* from its self-asserted level.' 逐字可查。
- [ok] 零脚本零代码（安全结论基础） — 目录 glob 仅 5 个 md 文件；token 扫描无任何执行/网络/写文件关键词。
- [ok] agent-reach / qcc / deep-research 调用点真实存在 — SKILL.md 三处明文（agent-reach 覆盖社媒平台、qcc 覆盖工商、deep-research 复用扇出模式）；均为外部依赖，不随本目录分发——已注明。
- [ok] 功能声明 vs 实际能力夸大检查 — description 承诺的『splitting marketing bubble from real signal, then mapping the validated playbook』全部由编排+分级+映射提示词兑现；无自动执行/自动写文件承诺，无越界。
- [ok] 元数据（commit/license/stars） — 本地 HEAD == pin d5c4678cb5d4fd6acc9c922690df035dbd33d247；GitHub API：MIT、stars 1385（2026-09-09 实采；候选池 2026-09-06 快照为 1378）、pushed_at 2026-09-09T12:33:29Z。
6Verdict & Scenarios
Best for: Ideal for users seeking to 'debunk, benchmark, or deconstruct' a founder, KOL, company, or product whose achievements appear inflated (growth leads, founders, investment researchers, content strategists), producing an action-oriented report that separates hype from replicable tactics and grounds them in the user's own resources. Also suitable as a pre-deal/competitor deep-dive for high-profile benchmarks. Requires users willing to run a multi-agent fan-out (higher latency and token consumption) and capable of providing authentic commissioner resource context.
Not for: Not suitable for scenarios requiring neutral, factual overviews (choose deep-research instead); not suitable for lightweight queries wanting only a quick public summary without running 12+ agents; not suitable when the commissioner cannot provide their own resource context (the mapping phase will fail) or when investigating private entities requiring specialized data sources (corporate registry or social media data dependent on external capabilities like qcc or agent-reach).
Install Agent-Ready
① Verified Mirror 2026-09-09
Option A · Direct DownloadDownload benchmark-due-diligence.tar.gz
sha256:
Option B · Copy Guide for Agenta3679e98097b1486…Agents automatically download and verify mirrors from this JSON guide.
② Upstream GitHub
GitHub Repo ↗
Pinned to commit d5c4678cb5.