Evaluate Technology

Rules for criteria:

• Add domain-specific criteria where relevant (e.g., "HIPAA compliance" for healthcare, "bundle size" for frontend)
• Remove criteria that genuinely don't apply — justify each removal
• Assign weights BEFORE research to prevent post-hoc rationalisation
• Every criterion must have a weight — no blanks

Output: A completed criteria table with weights and justifications.

Step 2: Research each option

For each technology under evaluation, produce a structured research brief:

### [Technology name]

**What:** [One-sentence description — what it is and what problem it solves]
**Version:** [Current stable version, release date]
**License:** [License type — MIT, Apache 2.0, BSL, proprietary, etc.]
**Notable adopters:** [3–5 companies or projects using it in production]

#### Maturity signals
- First stable release: [date]
- Release cadence: [frequency]
- Breaking changes in last 2 major versions: [count and severity]
- Open issues / PRs: [count, trend]

#### Community signals
- GitHub stars: [count] | npm/PyPI weekly downloads: [count]
- Stack Overflow questions: [count, answer rate]
- Active maintainers: [count, bus factor]
- Discord/forum activity: [qualitative — active, moderate, quiet]

#### Known limitations
- [Limitation 1 — sourced from issues, forums, or migration guides]
- [Limitation 2]
- [Search: "[tech] problems", "[tech] limitations", "[tech] vs alternatives"]

Output: One research brief per option.

Step 3: Score against criteria

Rate each option against each criterion on a 1–5 scale. Multiply by weight for a weighted score.

| Criterion | Weight | Option A | | Option B | |
|---|---|---|---|---|---|
| | | Raw (1–5) | Weighted | Raw (1–5) | Weighted |
| Maturity / stability | 5 | 4 | 20 | 3 | 15 |
| Community / ecosystem | 3 | 5 | 15 | 4 | 12 |
| ... | | | | | |
| **Total** | | | **X** | | **Y** |

Rules for scoring:

• Every score must have a one-sentence justification — no bare numbers
• If you lack data for a criterion, score it 3 (neutral) and flag it as "unverified"
• Do not round or average — the weighted total is the score

Output: Completed scoring matrix with justifications.

Step 4: Identify trade-offs and risks

No technology wins on every dimension. Explicitly state:

### Trade-offs

| Choosing Option A means... | Choosing Option B means... |
|---|---|
| [Advantage A has over B] | [Advantage B has over A] |
| [Cost of choosing A] | [Cost of choosing B] |

### Risks to monitor post-adoption

| Risk | Trigger signal | Mitigation |
|---|---|---|
| [Risk 1] | [What you'd see if this risk materialises] | [What you'd do] |
| [Risk 2] | [Trigger] | [Mitigation] |

Output: Trade-off comparison and risk register.

Step 5: Recommend

State the recommendation clearly:

### Recommendation

**Choose: [Option]**

**Primary reason:** [The single most important factor]
**Secondary reasons:** [Supporting factors]
**What we sacrifice:** [Explicit trade-off acknowledgement]
**Reconsideration triggers:** [Conditions that would change this recommendation]

If the evaluation warrants a formal record, suggest writing an ADR using /architect:write-adr.

If neither option is clearly better, say so — recommend a time-boxed spike or prototype instead of a forced choice.

Anti-Patterns (NEVER do these)

• Conclusion-first evaluation — deciding the winner before scoring, then fitting criteria to match. Weights must be set in Step 1, before research.
• Popularity as proxy — GitHub stars and npm downloads measure awareness, not quality. They are one signal among many, not the deciding factor.
• Ignoring total cost of ownership — a free library with no docs costs more than a paid one with good support. Score maintenance burden and team familiarity, not just the license.
• Comparing unequals — comparing a framework to a library, or a managed service to a self-hosted tool, without normalising the scope. State what each option requires you to build on top.
• Binary scoring — scoring everything 1 or 5 defeats the purpose. Use the full 1–5 range with justifications.
• Skipping the "what if we're wrong" step — every recommendation needs reconsideration triggers. Technology choices are expensive to reverse.

Output Format

# Technology Evaluation: [subject]

## Evaluation Criteria
[Weighted criteria table from Step 1]

## Research
### [Option A]
[Research brief]
### [Option B]
[Research brief]

## Scoring Matrix
[Weighted scoring table from Step 3]

## Trade-offs
[Trade-off table from Step 4]

## Risks
[Risk register from Step 4]

## Recommendation
[Recommendation from Step 5]

## Recommended Follow-ups
- [ ] ADR for this decision (if applicable)
- [ ] Spike/prototype for low-confidence areas
- [ ] Re-evaluate on [date or trigger condition]