quick

Cross-Lens: Quick

Fast single-discipline analysis. One question, one insight, one experiment. Under 3 minutes.

Tone: Same as spark — intellectually provocative, structurally rigorous, zero fluff.

---

PHASE 0: USER QUESTION FIRST

Quick question before I scan: what's bugging you most about this project right now?

Wait for answer. Store as USER_CONCERN.

---

PHASE 1: FAST SCAN (30 seconds max)

Step 1.1: Load References

Read from ${CLAUDE_PLUGIN_ROOT}/references/:

• scan-patterns.md
• domains.md

Step 1.2: Project Identity

Read README and package manifest only. Extract: what is this, who is it for.

Step 1.3: Domain Detection

Match codebase signals against domains.md. Select primary domain.

Step 1.4: Targeted Scan

Scan only the top 3 scan targets most relevant to USER_CONCERN:

• "growth/retention" → onboarding, pricing, feedback loops
• "complexity/debt" → dependencies, error handling, data flow
• "performance" → data flow, retry/resilience, feedback loops
• "adoption" → onboarding, configuration, error handling
• "general" → auth flow, error handling, configuration

For each: Glob to find files, Read top 2-3 files (first 50 lines), extract the key metric.

Step 1.5: Sparse Check

If fewer than 2 scan targets return results → use absence-based analysis.

Step 1.6: Select Single Discipline

Read all 5 principle files from ${CLAUDE_PLUGIN_ROOT}/references/principles/.

Simplified scoring — for each principle:

SCORE = SCAN_HIT (0-3) + CONCERN_MATCH (0-3)

Select the single discipline with the highest-scoring principle. From that discipline, select the principle with the highest score.

---

PHASE 2: SILENT ANALYSIS

Generate one principle-to-code mapping:

• Must reference a specific file, count, or absence
• Must be impossible to generate without codebase access
• Must include a concrete refactored approach

Generate one experiment:

• Directly derived from the mapping
• Include minimum viability measurement
• Include null result interpretation
• Directional predictions only (no round percentages)

---

PHASE 3: DELIVER

Output the quick report:

## Cross-Lens Quick: {Discipline Name}

### {Principle ID}: {Principle Name}

**What I found in your code:**
{Specific file/flow/pattern with quantitative detail}

**What {source author}'s {discipline} says about this:**
{The principle applied specifically to the finding}

**The insight:**
{One paragraph, specific, referencing code}

**Try this experiment:**
- **Change:** {exactly what to do}
- **Measure:** {exactly what metric}
- **Minimum viability measurement:** {how to measure without analytics}
- **Expect:** {directional prediction}
- **Timeline:** {how long to run it}
- **If nothing changes:** {null result interpretation}

---
*Run `/cross-lens:spark` for the full 5-discipline analysis with contradiction detection and dangerous ideas.*

---

ANTI-HOROSCOPE CHECK

Before outputting, verify:

1. The insight references a specific file or metric from the scan
2. It could NOT be generated without reading the codebase
3. The experiment has a minimum viability measurement
4. No round-number percentage predictions

If the insight fails any check, pick the next-highest-scoring principle and try again.