Skip to content

AI agent skills for startup validation, competitive intelligence, and planning

Add Radical Honesty Protocol, kill criteria, and expanded installation methods

Ferdinando Bonsegna919d824b91eb
diff --git a/README.md b/README.md
index 841a7ad..7502353 100644
--- a/README.md
+++ b/README.md
@@ -46,25 +46,65 @@ Each phase builds on the previous one. Sessions can be interrupted and resumed f
## Installation
-### Claude Code
+### Option 1: CLI Install (Recommended)
-Add as a plugin:
+Use [npx skills](https://github.com/vercel-labs/skills) to install skills directly:
```bash
-# From the Claude Code CLI
-/install-plugin ferdinandobons/startup-skill
+# Install the skill
+npx skills add ferdinandobons/startup-skill
+
+# List available skills
+npx skills add ferdinandobons/startup-skill --list
+```
+
+This automatically installs to your `.agents/skills/` directory (and symlinks into `.claude/skills/` for Claude Code compatibility).
+
+### Option 2: Claude Code Plugin
+
+Install via Claude Code's built-in plugin system:
+
+```bash
+# Add the plugin
+/plugin marketplace add ferdinandobons/startup-skill
+
+# Install
+/plugin install startup-design
```
-Or clone and use directly:
+### Option 3: Clone and Copy
+
+Clone the entire repo and copy the skill folder:
```bash
git clone https://github.com/ferdinandobons/startup-skill.git
-cd startup-skill
+cp -r startup-skill/startup-design .agents/skills/
+```
+
+### Option 4: Git Submodule
+
+Add as a submodule for easy updates:
+
+```bash
+git submodule add https://github.com/ferdinandobons/startup-skill.git .agents/startup-skill
```
-### Other Agents
+Then reference the skill from `.agents/startup-skill/startup-design/`.
+
+### Option 5: Fork and Customize
-Copy the `startup-design/` directory into your agent's skills folder. The skill follows the standard `SKILL.md` format with `references/` for supplementary documents.
+1. Fork this repository
+2. Customize the skill for your specific needs
+3. Clone your fork into your projects
+
+### Option 6: SkillKit (Multi-Agent)
+
+Use [SkillKit](https://github.com/rohitg00/skillkit) to install skills across multiple AI agents (Claude Code, Cursor, Copilot, etc.):
+
+```bash
+# Install the skill
+npx skillkit install ferdinandobons/startup-skill
+```
## Usage
diff --git a/startup-design/SKILL.md b/startup-design/SKILL.md
index b28064c..8468f4e 100644
--- a/startup-design/SKILL.md
+++ b/startup-design/SKILL.md
@@ -432,6 +432,10 @@ In `06-validation/experiment-design.md`:
- Hypothesis, method, metrics, success criteria, timeline
- Templates the founder can use immediately (interview scripts, survey questions, landing page copy outline)
+### Kill Criteria
+
+In `06-validation/kill-criteria.md`, define 5-7 specific, measurable conditions under which the founder should stop or pivot. Tie each to a validation experiment. Be specific: "If fewer than 3/10 interview subjects say they'd pay $X" not "if there's no demand." This protects the founder from sunk-cost thinking.
+
### Idea Scorecard
At the end of the validation section, produce a summary scorecard in `06-validation/scorecard.md`:
@@ -447,7 +451,7 @@ At the end of the validation section, produce a summary scorecard in `06-validat
| Timing | | |
| **Overall** | | |
-Be honest. If the idea has weaknesses, say so clearly. The goal is to help the founder make a good decision, not to validate their ego.
+Be honest. If the idea has weaknesses, say so clearly. The goal is to help the founder make a good decision, not to validate their ego. Include a clear **Verdict** paragraph after the table with an unambiguous recommendation (see the scoring guide in the Radical Honesty Protocol).
Update PROGRESS.md — mark all phases complete.
@@ -465,17 +469,45 @@ After all phases are complete, create a `README.md` at the project root that ser
---
-## Important Principles
+## Radical Honesty Protocol
+
+This skill exists to help founders make good decisions — not to feel good. An AI that cheerleads every idea is actively harmful. These principles are non-negotiable.
+
+### Tell the truth, even when it's uncomfortable
+
+- If the market is too small, say so directly. Don't soften "$12M and shrinking" into "room for a focused player."
+- If the idea has a fatal flaw, name it up front. Don't bury it in minor risks.
+- If the founder's assumptions contradict research, flag it: "You assumed X, but the data shows Y."
+- Challenge "everyone needs this" (who specifically?), "there's no competition" (there's always competition), and unsupported market claims.
+
+### Separate facts from opinions
+
+- Label claims: **[Data]** sourced findings, **[Estimate]** projections, **[Assumption]** unverified beliefs, **[Opinion]** your judgment.
+- When data is missing or weak, say so. A confident fabrication is worse than "I don't know."
+- Never present estimates as facts. "Revenue could reach $500K **assuming** 2% conversion" — not "Revenue will reach $500K."
+
+### Surface flags proactively
+
+In every phase output, include a **Flags** section:
+- **Red Flags** — Could kill the business (tiny market, dominant incumbent, broken unit economics, regulatory risk).
+- **Yellow Flags** — Needs investigation (high CAC, no domain expertise, requires behavior change, crowded market).
-1. **Be honest, not encouraging.** A skill that tells every founder their idea is great is worthless. Surface real risks, question weak assumptions, and point out when the market says "no." The most valuable thing you can do is help someone avoid spending a year on a doomed idea.
+If no flags exist for a phase, write "No flags identified" — don't skip the section.
-2. **Ground everything in evidence.** Every claim in the strategy and brand sections should trace back to something from the research phase. If you don't have data, say so explicitly.
+### Provide a clear verdict
-3. **Make it actionable.** Every document should leave the founder knowing what to do next. Avoid abstract frameworks that don't connect to concrete actions.
+The Phase 8 scorecard must include an unambiguous recommendation:
+- **8-10:** Strong proceed signal. Explain what's compelling.
+- **6-7:** Conditional proceed. List what must be validated first.
+- **4-5:** Significant concerns. Run experiments before committing.
+- **1-3:** Recommend against in current form. Suggest pivots if any show promise.
-4. **Respect the founder's time.** Don't generate filler. If a section isn't relevant to this particular startup, skip it and note why in PROGRESS.md.
+### Other principles
-5. **Track everything.** Always update PROGRESS.md after completing each phase. This is the user's lifeline if the session breaks.
+- **Ground in evidence.** Every strategy/brand/product claim should trace to research. No data? Say so.
+- **Make it actionable.** Every document should tell the founder what to do next.
+- **Respect the founder's time.** Skip irrelevant sections; note why in PROGRESS.md.
+- **Track everything.** Update PROGRESS.md after each phase — it's the session lifeline.
---