teardown

name: teardown description: "Use when the user wants a deep dive into a specific evaluation dimension for a product. Example: teardown moat, teardown journey. Goes deeper than the standard evaluation on one dimension."

Process

Step 1: Parse Dimension

Accept the dimension argument in both Chinese and English. Mapping:

Also accept partial matches and common aliases:

• need / needs / demand / jtbd → Demand Authenticity
• loop / journey / flow / ux → Journey Completeness
• market / space / competition → Market Space
• business / revenue / money / viability → Business Viability
• moat / defensibility / sherlock → Moat
• execution / quality / tech / debt → Execution Quality

If the argument doesn't match any dimension, list the available options and ask the user to choose.

Step 2: Determine Target

Same logic as evaluate skill:

• Path argument → local project
• Name/URL argument → external app
• No argument → current working directory

Step 3: Detect Platform and Resolve Paths

1. Detect platform (iOS / Web / etc.)
2. Locate reference files by searching for **/product-lens/references/_calibration.md. From the same parent directory, resolve the path to:
   • _calibration.md
   • The single dimension file matching the target dimension (from the mapping in Step 1)

Step 4: Pre-merge Sub-Questions

Read _calibration.md (preamble for the evaluator).

Read the target dimension file. Extract:

1. Core question
2. Universal sub-questions
3. Platform-specific sub-questions (from ### iOS if iOS detected, otherwise ### Default)
4. Scoring anchors
5. Evidence sources

If iOS detected and the dimension has an iOS core question variant (blockquote under ### iOS), note it for the evaluator.

Merge universal + platform-specific into a single numbered list.

Step 5: Gather Market Context (if applicable)

For dimensions that benefit from market data (Demand Authenticity, Market Space, Business Viability, Moat):

• Dispatch market-scanner agent with product info
• Wait for completion

For other dimensions (Journey Completeness, Execution Quality):

• Skip market-scanner; these are primarily code/product analysis

Step 6: Deep Evaluation

Dispatch a single dimension-evaluator agent with:

• Calibration context: full content of _calibration.md
• Dimension name (English + Chinese) and core question
• Sub-questions: the merged numbered list from Step 4
• Scoring anchors from Step 4
• Evidence source hints from Step 4
• Product info: name, description, type, project root
• Market data: from Step 5 (or "none")
• Depth mode: deep

Wait for completion.

Step 7: Present Results

Display the deep-dive report from the dimension-evaluator. The deep mode output includes:

• Per-sub-question analysis with sub-scores
• Evidence summary table
• Dimension score with anchor match
• Prioritized recommendations
• Related dimensions observations

Post-processing:

• If the dimension scored <=2, highlight this prominently
• If Related Dimensions section identifies signals for other dimensions, suggest running /evaluate for the full picture or /teardown [other dimension] for a focused follow-up

Completion Criteria

• Target dimension identified and deep-evaluated (depth mode: deep)
• Deep-dive report with per-sub-question analysis presented
• Follow-up suggestions provided (if related dimensions have signals)