Comparative Analysis

Build weighted comparison matrices across options.

Skill metadata

Source	Bundled (installed by default)
Path	skills/research/comparative-analysis
Version	1.0.0
Author	VibeOS
License	MIT
Platforms	linux, macos, windows
Tags	research, analysis, comparison, matrix, decision
Related skills	professional-research, source-critique

Reference: full SKILL.md

info

The following is the complete skill definition that VibeOS loads when this skill is triggered. This is what the agent sees as instructions when the skill is active.

Comparative Analysis Skill

Compare options (vendors, approaches, hypotheses) in a weighted matrix with cited evidence — not a bullet list of opinions.

When to Use

• “Compare A vs B vs C”, vendor selection, architecture tradeoffs, market scan.
• After gathering sources via professional-research / web_search.

Prerequisites

• Clear decision question and candidate set (ask if missing).
• Evidence from search/extract/MCP; critique weak sources first if contested.

How to Run

1. Restate the decision and constraints (budget, region, deadline, must-haves).
2. Agree criteria (3–8). Assign weights that sum to 100.
3. Score each option 0–5 per criterion with a citation.
4. Compute weighted totals; state recommendation + risks.

Helper (optional) — from this skill directory:

python3 scripts/build_matrix.py <<'EOF'
{
  "criteria": [
    {"id": "cost", "weight": 30},
    {"id": "quality", "weight": 40},
    {"id": "risk", "weight": 30}
  ],
  "options": [
    {
      "name": "Option A",
      "scores": {"cost": 4, "quality": 3, "risk": 4},
      "notes": {"cost": "cite [1]", "quality": "cite [2]", "risk": "cite [3]"}
    }
  ]
}
EOF

Output template

# Comparison: <decision>

## Question & constraints
...

## Criteria (weights = 100)
| Criterion | Weight | Why it matters |
|-----------|--------|----------------|

## Matrix
| Criterion (w) | Option A | Option B | Option C |
|---------------|----------|----------|----------|
| ... (w) | score — evidence [n] | ... | ... |
| **Weighted total** | **x.x** | ... | ... |

## Recommendation
- Pick: ...
- Why: ...
- Risks / what would change the answer: ...

## Sources
1. ...

Scoring rules

• 0 = fails / unknown after search; 5 = best among candidates on that criterion.
• Unknown after honest search → score 0–1 and mark “unknown”, don’t invent.
• Every cell with score ≥ 3 needs a citation.
• Invert scales consistently (e.g. lower cost → higher score).

Pitfalls

• Don’t average unrelated criteria without weights.
• Don’t declare a winner when top two totals are within ~5% without caveats.
• Don’t mix marketing claims and measured data without labels.

Verification

• Weights sum to 100.
• Every high score has a citation.
• Recommendation states what would overturn it.