# nerissa-self-audit
> **Stop only running rigorous frameworks on your competitors.** A disciplined Claude Skill that turns 9 classic strategy lenses (First Principles + JTBD + 5 Whys + OKR + SWOT + Porter + BCG + Blue Ocean ERRC + Inversion) inward — on your own startup, with the same honesty you'd use to analyze a competitor you're skeptical of.
[](LICENSE)
[](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview)
---
## The Problem
Most founders happily run SWOT, Porter Five Forces, JTBD, and Inversion on competitors — but never on themselves. The result:
1. **Vision-vs-reality gaps** — your landing page says "for X" but every signed customer is "Y"
2. **80% of team time on Dogs** — features that score low on both growth and traction, but no one has the heart to kill them
3. **Wrong competitor framing** — you benchmark against $1B-funded giants, but your real competitor is "doing nothing"
4. **5 layers of root cause untouched** — you treat every symptom (no sales, high churn) without asking "why" 5 times
5. **Death scenarios in progress** — Charlie Munger said "invert, always invert" — most teams never list the ways they'd certainly die
Each of these is a documented failure mode this skill prevents.
## What This Skill Does Differently
It forces you to:
- **Run all 9 lenses, not just the comfortable ones** — cherry-picking SWOT because it makes you look good is the most common mistake
- **Run each lens independently before synthesizing** — letting your favorite conclusion bias each lens defeats the purpose
- **Look for convergence across lenses** — if 3+ lenses point to the same diagnosis, that's signal; if only 1, it's noise
- **Apply same scrutiny as to a competitor** — first-person voice ("we are") but external-style honesty
- **Cite evidence per lens** — not vibes · data, quotes, code paths
## Quick Start
### Use as a Claude Code Skill
1. Clone this repo into your Claude skills directory:
```bash
mkdir -p ~/.claude/skills
cd ~/.claude/skills
git clone https://github.com/xuenerissa-png/nerissa-self-audit.git
```
2. In any Claude Code session, trigger by saying:
- `"audit our current state"`
- `"self-audit using the 9-lens skill"`
- `"我们到底是什么 · 用 self-audit"`
- `"where are we really · run the 9 lenses"`
3. Claude will walk through all 9 lenses, find convergence, and produce diagnosis + decisions blocking progress.
### Use without Claude (as a manual workshop)
The `SKILL.md` + `references/9-lens-worksheet.md` form a self-contained 2-hour workshop. Print, work through alone or with team, fill the convergence map, surface diagnoses.
## The 9 Lenses (Quick Summary)
1. **First Principles** — strip assumptions, find atomic facts
2. **JTBD** — what job are customers actually hiring you for?
3. **5 Whys** — drill from symptom to root cause
4. **OKR** — what would prove the next quarter worked?
5. **SWOT** — internal strengths/weaknesses × external opportunities/threats
6. **Porter Five Forces** — what shapes industry profitability?
7. **BCG Matrix** — which features are Stars / Cash Cows / Question Marks / Dogs?
8. **Blue Ocean ERRC** — Eliminate / Reduce / Raise / Create
9. **Inversion** — how would you certainly fail? (then avoid)
## The Critical Synthesis Step
After running all 9 lenses independently, ask:
> **Which 1-2 conclusions did ALL OR MOST lenses point to?**
That convergence is your real diagnosis. Single-lens insights are lower confidence. The skill enforces this discipline.
## Output Format
For each audit session, you get:
```
🔬 Lens 1 · First Principles
[atomic facts + identity re-derivation]
🎯 Lens 2 · JTBD
[real customer job + real competitor]
❓ Lens 3 · 5 Whys
[symptom → root cause chain]
📐 Lens 4 · OKR
[O + 3-5 KRs for next quarter]
🧭 Lens 5 · SWOT
[4-quadrant matrix + Weakness×Opportunity blocker]
⚔️ Lens 6 · Porter Five Forces
[5 forces scored + main threat]
📊 Lens 7 · BCG Matrix
[features by quadrant + % time on Dogs]
🌊 Lens 8 · Blue Ocean ERRC
[4 actions for category redesign]
🔁 Lens 9 · Inversion
[5-10 death scenarios + which are in progress]
🎯 CONVERGENCE MAP
[which conclusions appeared in 3+ lenses]
📋 TOP 1-3 REAL DIAGNOSES
❓ 5 REAL QUESTIONS ONLY YOU CAN ANSWER
🚫 ANTI-FLATTERY FLAGS
[places where the audit was making things look better than reality]
```
## Examples
5 worked examples in `examples/` (fully anonymized):
- **01 · Vision-vs-Reality Misalignment** — Landing page says "for solo founders" but 0 actual signed customers fit that profile; all signed customers are mid-market B2B. First Principles + JTBD + SWOT all converge on "your vision and your real customers are 4 dimensions apart."
- **02 · Resource Misallocation via BCG** — Team thinks they're spending 50/50 on core product vs experimental features. BCG matrix reveals 80% of recent commits are on features that score as "Dogs" (low growth + low share). Killing them frees 40% of engineering time.
- **03 · Wrong Competitor via JTBD** — Team benchmarks against well-funded AI startups. JTBD reveals customers hired them over "manually doing it themselves" — meaning real competitor is "status quo + outsourced agencies." Pitch deck rebuilt around this.
- **04 · 5 Whys Finds the Real Blocker** — Symptom: no signed customers in 30 days. 5 layers down, root cause = "we never decided on final pricing tiers, so the sales team can't send a quote." The blocker wasn't sales execution — it was an undecided strategic input.
- **05 · Inversion Catches the Quiet Death** — Team listed 5 ways the startup would die. Discovered #3 ("we keep researching competitors but never sign customers") was already happening for 6 weeks. Stopped immediately, redirected to closing.
## Why This Exists
I built this while running a competitor research practice for an early-stage startup. I noticed I was applying rigorous frameworks to other companies but never applying the same rigor to my own situation. When I finally did — using the 9 lenses I'd already been using on competitors — I found embarrassing convergence:
- The product had been built for a customer profile that didn't match the actual signed pipeline
- 80% of recent feature work was on functionality that scored as Dogs
- The team had been comparing to $1B-funded giants when our real competitor was "customers doing nothing"
- The actual blocker to signing customers was an undecided pricing question, not anything in the product
Each finding came from a different lens, but they all converged on the same diagnosis: **the strategy and operations were out of sync, and no single framework would have shown it.**
After that, the self-audit became a quarterly practice. This skill makes it repeatable.
## Comparison to Other Approaches
| Approach | What it gives you | What it misses |
|---|---|---|
| **"Strategy off-site" once a year** | Big-picture re-think | Quarterly drift between strategy and reality |
| **Single framework (just SWOT, just OKR)** | One angle | Misses convergence across frameworks |
| **Investor update writing** | Forces clarity | Optimization for narrative, not honesty |
| **nerissa-self-audit** | 9 lenses + convergence map + anti-flattery discipline | More upfront work (2 hours) |
If you do strategic decisions less than once per quarter, you probably don't need this. If you face strategic decisions every 4-6 weeks, the convergence discipline pays back fast.
## Roadmap
- [x] v1.0 · 9-lens core + convergence map + 5 examples + worksheet
- [ ] v1.1 · Industry-specific lens variants (consumer SaaS / B2B / marketplace / hardware)
- [ ] v1.2 · Audit-comparison-over-time (so quarterly drift is visible)
- [ ] v1.3 · Pre-fundraise audit variant (with investor-readiness scoring)
## Related Skills
- [decision-lens](https://github.com/xuenerissa-png/decision-lens) — Multi-role pressure-test for ONE specific decision (this skill audits the whole startup state)
- [nerissa-competitive-research](https://github.com/xuenerissa-png/nerissa-competitive-research) — Apply the same rigor to OTHER companies (this skill turns it inward)
The three form a triangle: decision-lens for single decisions, competitive-research for external, self-audit for internal.
## License
MIT · Free to use, fork, adapt.
## Author
Built by [@xuenerissa-png](https://github.com/xuenerissa-png). If you find this useful or want to contribute a new convergence pattern, file an issue.
---
中文版 README: [README_zh-CN.md](README_zh-CN.md)