feat: add unblock skill to diagnose and resolve analysis paralysis

.claude/settings.json

@@ -50,6 +50,8 @@
       "Skill(scratch-review)",
       "Skill(idea-refine)",
       "Skill(notes-sync)",
+      "Skill(unblock)",
+      "AskUserQuestion",
       "WebSearch",
       "mcp__context7__get-library-docs",
       "mcp__context7__resolve-library-id",

.claude/skills/unblock/SKILL.md

@@ -0,0 +1,100 @@
+---
+name: unblock
+description: Use when user says "unblock", "am I overthinking", "help me ship", "ship check". Diagnostic tool using Yes/No questions to identify if caution is justified or paranoia. Provides actionable prescriptions.
+allowed-tools: AskUserQuestion
+---
+
+# unblock
+
+Break analysis paralysis by diagnosing if caution is justified or paranoia.
+
+## Process
+
+### 1. Context (via AskUserQuestion tool)
+
+Analyze recent session to infer what's happening, then use AskUserQuestion to confirm:
+
+**Step 1**: Review last 10-20 messages for patterns:
+- Repeated refactoring/rewrites
+- Long research/planning without implementation
+- Working on meta-tools instead of core features
+- Multiple "what if" or "should I" questions
+
+**Step 2**: Use AskUserQuestion with inferred context:
+```
+question: "What are you stuck on?"
+options:
+  - [Inferred from session, e.g. "Perfecting the unblock skill for 1+ hour"]
+  - [Alternative inference if applicable]
+  - "Something else entirely"
+```
+
+If time unclear, follow with:
+```
+question: "How long on this without shipping?"
+options: ["< 30 mins", "30m - 2h", "2h - 1 day", "> 1 day"]
+```
+
+### 2. Diagnostic (15-20 questions via AskUserQuestion tool)
+
+<critical>
+ALL questions MUST use AskUserQuestion tool - including initial context.
+Interactive buttons prevent overthinking and force decisions.
+</critical>
+
+Ask in 3-4 batches of 4-5 questions. Adapt based on context.
+
+**Batch 1: Risk & Velocity**
+- Can users lose data/money? (Stakes)
+- Shipped last 2 weeks? (Velocity)
+- >2 hours researching? (Analysis)
+- Reputation damage worry? (Fear)
+
+**Batch 2: Users & Demand**
+- Anyone asked for this? (Demand)
+- Users complain if nothing ships? (Pressure)
+- Solving real problem? (Validation)
+- Rewritten without feedback? (Iteration)
+
+**Batch 3: Complexity & Focus**
+- Ship with half the code? (Scope)
+- >3 active blockers? (Focus)
+- Complexity for "professionalism"? (Theater)
+- More thinking than coding? (Action)
+
+**Batch 4: pgflow & Momentum**
+- Critical onboarding path? (Priority)
+- Treating MVP like enterprise? (Scale)
+- Fear from actual feedback? (Reality)
+- Know NOT shipping kills projects? (Truth)
+
+### 3. Score
+
+Count flags:
+- **RED**: Low stakes + high fear + no users = Paranoia
+- **YELLOW**: Mixed signals
+- **GREEN**: Real stakes + users waiting = Justified
+
+### 4. Diagnose
+
+```
+## Diagnosis: [Paranoia/Mixed/Justified]
+
+### Evidence:
+- RED: [specific findings]
+- GREEN: [what works]
+
+### Problem: [1 paragraph root cause]
+
+### Cost: [Quantified loss]
+
+### Prescription: [See prescriptions.md]
+```
+
+### 5. Follow-up
+"Want help shipping the blocked thing now?"
+
+## Resources
+
+- [patterns.md](patterns.md) - Pattern detection
+- [prescriptions.md](prescriptions.md) - Treatments by type

.claude/skills/unblock/patterns.md

@@ -0,0 +1,19 @@
+# Patterns
+
+## Red Flags (Paranoia)
+- **Analysis Paralysis**: >2h research, multiple rewrites, >5 blockers
+- **Impostor Syndrome**: Fear of exposure, zero complaints but worried
+- **Momentum Paranoia**: "One bug kills project", no actual negative feedback
+- **Complexity Theater**: Over-engineering to appear professional
+- **Scope Creep**: "While I'm at it" additions
+
+## Green Flags (Justified)
+- **Real Stakes**: Paying customers, data loss risk, financial transactions
+- **Active Users**: Users asked for it, complaints about delays
+- **pgflow Critical Path**: Install/compile/onboarding must work
+
+## Quick Detection
+- No paying customers + high fear = Paranoia
+- <50 users + scale worries = Premature optimization
+- Zero complaints + embarrassment = Impostor syndrome
+- Rewriting without shipping = Perfectionism

.claude/skills/unblock/prescriptions.md

@@ -0,0 +1,45 @@
+# Prescriptions
+
+## High Paranoia → Radical Action
+1. Ship in 2 hours max (set timer)
+2. Delete 4 of 5 blockers
+3. 30-min research cap
+4. Can't research what users didn't ask for
+
+*Ship the thing. Fix what breaks. Repeat daily.*
+
+## Impostor Syndrome → Build in Public
+1. Ship with "🚧 Experimental" prefix
+2. Add "Known Limitations" to README
+3. Ask for feedback explicitly
+4. Make reporting issues easy
+
+*Your 177 stars want to help shape, not judge.*
+
+## Complexity Theater → Aggressive Simplification
+1. Delete 50% of code
+2. Remove ALL future abstractions
+3. Ship simplest solution that works
+
+*Complexity is debt. Ship simple.*
+
+## pgflow Onboarding → First Experience Priority
+```bash
+npx pgflow install  # MUST work
+npx pgflow compile  # MUST work
+```
+*Silent abandonment is real. Protect onboarding.*
+
+## Justified Caution → Controlled Risk
+1. Ship behind feature flag
+2. Test with friendlies first
+3. Progressive rollout over 2 weeks
+
+*Real stakes need real process.*
+
+## Mixed Signals → Time-boxed Polish
+1. List concerns (10 min)
+2. Fix top 2 user-affecting (4h max)
+3. Ship with "Beta" label
+
+*Good enough beats perfect later.*