opc-checkup: business health check
There's only one subject: your own idea or running project. Deconstructing someone else's benchmark is opc-analysis's job. This skill turns the knife on yourself, to decide whether to build it, keep going, or change course.
Stance discipline: confirmation bias is the biggest risk when it's your own project. Run the checkup adversarially: the job is to find the reasons it should die, and it only passes if you can't find any. You're not its defense attorney.
The seven-step checkup
The spine of the seven steps: 1-2 nail down the facts first, 3-5 make the judgment calls, 6 checks your own bias, 7 lands on action and verification.
- Cash-flow razor first: strip the narrative halo, then ask: can this business survive on real cash flow, or can you point to a clear path where the numbers work? Answer in one line, and let that answer decide where to dig hardest next.
- Three checklists (full tables in
references/checks.md): the business-logic checklist (the five factors, one by one) → the business-narrative checklist (seven dimensions) → the capability-circle and resource-constraint checklist (five sections). Work through every item, don't skip a table.
- Business lifespan pricing and typing: is this a short-window arbitrage play or a long-line compounding play? Four pricing questions: how long is the window (as a multiple of the payback period)? How many times over does the in-window return pay back? What does exit cost if the window is misjudged or closes? What transferable assets remain at the end? Compounding is a bonus, not a pass line, and a short window is no mark against you. But each type carries its own risk control: arbitrage types get checked for exit discipline and window misjudgment, compounding types get checked for bad investment (busy on paper with no asset accumulation).
- Quick unit-economics math: does LTV > CAC + COC hold? Rough out three ledgers (acquisition / fulfillment / fixed costs). LTV < CAC is a death spiral: the bigger the scale, the faster the death. Negative unit economics only earns an early exemption when all three conditions are met together: "paradigm-shift window + strong narrative capital + a nameable inflection point." Miss even one and it's a red flag. Anchor definitions: a paradigm-shift window means the cost or capability curve of the category is moving by an order of magnitude and you're on the side that benefits. Strong narrative capital means funding already raised, significant media coverage, or a waitlist with real scale. None of the three present means it doesn't count. A nameable inflection point means you can name the specific event, date, and number path. "It's coming soon" doesn't count.
- Fatal dependencies and antifragility: SPOF check: who could pull the ladder out and kill you (a single platform, a major customer, an upstream API, a key person)? How long could you last if the real worst case hit? Is there a barbell structure (most assets in a safe zone plus a small slice of high-volatility experiments)? If this failed, what would be left to take with you (skills, code, content, relationships, insight)?
- Bias-firewall self-check before concluding: whoever runs the checkup is naturally biased toward their own project, so go through every item: sunk cost (zero out what's already been invested and ask again: starting from zero today, would I still build this? "I've already put in N months" is never a reason to keep going on its own). Endowment effect (swap the project's name for a stranger's project: would the three checklists still score the same?). Planning fallacy (double the payback period and cost estimates: does the unit economics still hold?). Confirmation bias (look back at this checkup: did you reverse at least one earlier judgment? If not even once, you were probably going through the motions).
- An uncertainty-elimination plan: turn the biggest uncertainty the checkup surfaced into a minimal validation (one build → measure → learn loop). An MVP is a tool for eliminating uncertainty, not an excuse for shipping something unfinished. Output "what to falsify next round, how to falsify it, and what it will cost."
Output
Save the checkup report to the workspace draft directory .issues/<YYYY-MM-DD-topic>/checkup.md. The minimal validation plan and the recheck reminder both need somewhere to land. Before saving, make sure the draft directory is git-ignored: if git check-ignore .issues fails, append .issues/ to .git/info/exclude (a local ignore, leave the shared .gitignore alone) and give the user a one-line heads-up. In a non-git environment there's no commit risk, so just write it directly. Report structure: the cash-flow razor verdict / conclusions for each of the three tables, section by section (marked red/yellow/green: red means this item alone is enough to rule out continuing, yellow means it's in doubt and needs verification and must come with "what evidence would turn it green," and green means it's backed by evidence. Yellow is not a safety valve.) / lifespan typing and its matching risk controls / the unit-economics verdict / the SPOF list / four possible conclusions: continue (arbitrage types attach exit discipline, compounding types attach an asset-accumulation plan), continue with changes (spell out what changes), pivot (where the capability-circle assets go), or stop (loss-cutting steps and what's portable) / the bias-firewall self-check record (which one nearly fooled you) / the minimal validation plan (when the conclusion is stop, the validation plan is limited to a final falsification gate before cutting losses: tied to a hard deadline, run in parallel with loss-cutting prep, and never used as an excuse to delay cutting losses) / a reading prescription: for the one or two weakest factors the checkup exposed, prescribe 1-2 books and explain why. When several items are weak, prioritize by root cause: first work out which factor is the root cause and which are just its downstream symptoms, and don't prescribe books for the symptoms. Don't hand out a full reading list, just patch the shortest plank.
Hard rules
- Unverified external facts (market data, competitors, costs) must be flagged, and order-of-magnitude estimates count as fabrication too.
- A framework is a lens, not truth. State conclusions in probabilistic language.
- Boundaries with neighboring skills: if the direction itself is still in doubt (what to build hasn't been chosen yet) → validate the direction with opc-niche first, then come back for a checkup. If what needs deconstructing is someone else's product or person → opc-analysis.
1---2name: opc-checkup3description: 商业体检：给自己的 idea 或在营项目做理性体检，现金流 Razor、三张检查表、生意寿命定价、单元经济与致命依赖排查，输出分型结论。Business health check for your own idea or running project. Use when the user wants to 给 idea 做体检 / 这个项目还要不要做 / 该不该继续或停掉 / health-check my idea / should I keep building this. Do not trigger for deconstructing other people's products or founders（那是 opc-analysis 的事）.4---56# opc-checkup: business health check78There's only one subject: **your own idea or running project**. Deconstructing someone else's benchmark is opc-analysis's job. This skill turns the knife on yourself, to decide whether to build it, keep going, or change course.910**Stance discipline**: confirmation bias is the biggest risk when it's your own project. Run the checkup adversarially: the job is to find the reasons it should die, and it only passes if you can't find any. You're not its defense attorney.1112## The seven-step checkup1314The spine of the seven steps: 1-2 nail down the facts first, 3-5 make the judgment calls, 6 checks your own bias, 7 lands on action and verification.15161. **Cash-flow razor first**: strip the narrative halo, then ask: can this business survive on real cash flow, or can you point to a clear path where the numbers work? Answer in one line, and let that answer decide where to dig hardest next.172. **Three checklists** (full tables in `references/checks.md`): the business-logic checklist (the five factors, one by one) → the business-narrative checklist (seven dimensions) → the capability-circle and resource-constraint checklist (five sections). Work through every item, don't skip a table.183. **Business lifespan pricing and typing**: is this a short-window arbitrage play or a long-line compounding play? Four pricing questions: how long is the window (as a multiple of the payback period)? How many times over does the in-window return pay back? What does exit cost if the window is misjudged or closes? What transferable assets remain at the end? Compounding is a bonus, not a pass line, and a short window is no mark against you. But each type carries its own risk control: arbitrage types get checked for exit discipline and window misjudgment, compounding types get checked for bad investment (busy on paper with no asset accumulation).194. **Quick unit-economics math**: does LTV > CAC + COC hold? Rough out three ledgers (acquisition / fulfillment / fixed costs). LTV < CAC is a death spiral: the bigger the scale, the faster the death. Negative unit economics only earns an early exemption when all three conditions are met together: "paradigm-shift window + strong narrative capital + a nameable inflection point." Miss even one and it's a red flag. Anchor definitions: a paradigm-shift window means the cost or capability curve of the category is moving by an order of magnitude and you're on the side that benefits. Strong narrative capital means funding already raised, significant media coverage, or a waitlist with real scale. None of the three present means it doesn't count. A nameable inflection point means you can name the specific event, date, and number path. "It's coming soon" doesn't count.205. **Fatal dependencies and antifragility**: SPOF check: who could pull the ladder out and kill you (a single platform, a major customer, an upstream API, a key person)? How long could you last if the real worst case hit? Is there a barbell structure (most assets in a safe zone plus a small slice of high-volatility experiments)? If this failed, what would be left to take with you (skills, code, content, relationships, insight)?216. **Bias-firewall self-check before concluding**: whoever runs the checkup is naturally biased toward their own project, so go through every item: **sunk cost** (zero out what's already been invested and ask again: starting from zero today, would I still build this? "I've already put in N months" is never a reason to keep going on its own). **Endowment effect** (swap the project's name for a stranger's project: would the three checklists still score the same?). **Planning fallacy** (double the payback period and cost estimates: does the unit economics still hold?). **Confirmation bias** (look back at this checkup: did you reverse at least one earlier judgment? If not even once, you were probably going through the motions).227. **An uncertainty-elimination plan**: turn the biggest uncertainty the checkup surfaced into a minimal validation (one build → measure → learn loop). An MVP is a tool for eliminating uncertainty, not an excuse for shipping something unfinished. Output "what to falsify next round, how to falsify it, and what it will cost."2324## Output2526Save the checkup report to the workspace draft directory `.issues/<YYYY-MM-DD-topic>/checkup.md`. The minimal validation plan and the recheck reminder both need somewhere to land. Before saving, make sure the draft directory is git-ignored: if `git check-ignore .issues` fails, append `.issues/` to `.git/info/exclude` (a local ignore, leave the shared `.gitignore` alone) and give the user a one-line heads-up. In a non-git environment there's no commit risk, so just write it directly. Report structure: the cash-flow razor verdict / conclusions for each of the three tables, section by section (marked red/yellow/green: red means this item alone is enough to rule out continuing, yellow means it's in doubt and needs verification and must come with "what evidence would turn it green," and green means it's backed by evidence. Yellow is not a safety valve.) / lifespan typing and its matching risk controls / the unit-economics verdict / the SPOF list / **four possible conclusions: continue (arbitrage types attach exit discipline, compounding types attach an asset-accumulation plan), continue with changes (spell out what changes), pivot (where the capability-circle assets go), or stop (loss-cutting steps and what's portable)** / the bias-firewall self-check record (which one nearly fooled you) / the minimal validation plan (**when the conclusion is stop**, the validation plan is limited to a final falsification gate before cutting losses: tied to a hard deadline, run in parallel with loss-cutting prep, and never used as an excuse to delay cutting losses) / **a reading prescription**: for the one or two weakest factors the checkup exposed, prescribe 1-2 books and explain why. When several items are weak, prioritize by root cause: first work out which factor is the root cause and which are just its downstream symptoms, and don't prescribe books for the symptoms. Don't hand out a full reading list, just patch the shortest plank.2728## Hard rules2930- Unverified external facts (market data, competitors, costs) must be flagged, and order-of-magnitude estimates count as fabrication too.31- A framework is a lens, not truth. State conclusions in probabilistic language.32- Boundaries with neighboring skills: if the direction itself is still in doubt (what to build hasn't been chosen yet) → validate the direction with opc-niche first, then come back for a checkup. If what needs deconstructing is someone else's product or person → opc-analysis.
liustack