explore-alternatives

Explore Alternatives

When to invoke

BEFORE committing to ANY decision with complexity ≥5 (per supervibe:requirements-intake complexity score). Specifically:

• Library / framework choice
• Pattern adoption (when ≥2 patterns plausibly fit)
• Architecture decision
• Performance optimization approach (always profile-then-explore)
• Bug fix when root cause has ≥2 possible interventions

NOT for: trivial fixes, single-obvious-solution tasks.

This skill bans "first idea wins" thinking. Forces comparison.

Expert Operating Standard

Follow docs/references/skill-expert-operating-standard.md: start from source of truth, preserve retrieval evidence, apply scope safety, use real producers with runtime receipts for durable delegated outputs, verify before completion claims, and keep confidence below gate when evidence is partial.

Step 0 — Read source of truth (required)

1. Read task context (spec / plan / bug report)
2. Read supervibe:project-memory for prior decisions in this area
3. Check if existing PRD decision section or pattern already addresses (don't reinvent)

Decision tree

How many alternatives realistic?
├─ 0 (no alternatives possible) → STOP, this skill not needed; document why single path
├─ 1 (only one viable path) → document why others rejected; this skill done
└─ ≥2 → continue with full procedure

Source of alternatives:
├─ Domain knowledge (agent's training)
├─ Project memory (supervibe:project-memory)
├─ Research (best-practices-researcher / infra-pattern-researcher / dependency-researcher)
└─ User suggestion

Procedure

1. Step 0 — context + memory check
2. Brainstorm alternatives — minimum 2, ideally 3
3. For EACH alternative:
   • Description (1-2 sentences what it is)
   • Pros (3+ concrete advantages)
   • Cons (3+ concrete disadvantages)
   • Cost (effort to implement, ongoing maintenance, runtime cost)
   • Reversibility (easy to change later? expensive lock-in?)
   • When-to-use (which condition makes this best)
4. Comparison table:

   | Criterion | Alt A | Alt B | Alt C |
   |-----------|-------|-------|-------|
   | Effort    | High  | Low   | Med   |
   | Lock-in   | Low   | High  | Med   |
   | Perf      | Best  | OK    | Best  |
   | DX        | OK    | Best  | OK    |
   

5. Recommendation — explicit choice with rationale citing project context (constraints, prior decisions, team skills)
6. Score with supervibe:confidence-scoring (agent-output ≥9)
7. If decision is structural → propose supervibe:prd to record permanently
8. If user-facing decision → seek user approval before implementing

Output contract

## Alternatives Considered: <decision topic>

### Alt A: <name>
- Description: ...
- Pros: 1) ... 2) ... 3) ...
- Cons: 1) ... 2) ... 3) ...
- Cost: ...
- Reversibility: easy | medium | locked-in
- When best: ...

### Alt B: <name>
...

### Alt C: <name>
...

### Comparison
<table>

### Recommendation
**Chose: <Alt X>**
Rationale: <2-3 sentences citing project context>

### Next step
- PRD decision section: <link to .supervibe/artifacts/prd/NNNN.md if structural>
- OR: continue to writing-plans with chosen alternative

Guard rails

• DO NOT: skip this skill on complexity ≥5 decisions
• DO NOT: present 2 alternatives where one is obviously inferior (straw man)
• DO NOT: choose without explicit rationale citing context
• DO NOT: ignore project memory of prior similar decisions (consistency matters)
• ALWAYS: cite which constraint/context drove the choice
• ALWAYS: document when reversal would be cheap vs expensive

Verification

• ≥2 alternatives with all required fields
• Comparison table present
• Recommendation has explicit rationale
• Confidence ≥9

Related

• supervibe:project-memory — for prior similar decisions
• supervibe:prd — to record structural decisions
• supervibe:brainstorming — already includes "propose 2-3 approaches"; this skill is for tactical decisions DURING execution where brainstorming is overkill
• agents/_core/architect-reviewer — invokes this skill for architectural decisions

Carbon-copy lookup (mandatory pre-step)

BEFORE generating original alternatives, ask: has someone else solved this problem?

1. Invoke supervibe:project-memory with the problem keywords — past decisions in this repo
2. Invoke supervibe:code-search semantic — similar code patterns in the codebase
3. Invoke supervibe:_ops:best-practices-researcher if applicable — industry references
4. Invoke supervibe:_ops:competitive-design-researcher for design problems

If carbon copies exist: list them BEFORE generating new options. Often one of them is the answer.

Weighted decision matrix (required)

Required format:

| Dimension | Weight | A | B | C |
|-----------|--------|---|---|---|
| <name>    | <int>  | 0-10 | 0-10 | 0-10 |
| ...       | ...    | ... | ... | ... |
| **Total** | --     | (sum w_i × a_i) | ... | ... |

Rules:

1. Weights set BEFORE seeing options (prevents post-hoc rationalization)
2. Negative weights allowed for "less is better" dimensions (effort, risk)
3. At least 4 dimensions — fewer means lazy thinking
4. Show calculation — make math visible

Sensitivity analysis

After scoring, perturb the weights:

• "If I doubled the weight on Risk, would the winner change?"
• "If I halved the weight on Strategic Fit, would the runner-up win?"

If small weight changes flip the result: the matrix is brittle; revisit weights or add more dimensions.

Adversarial scoring

Force yourself to argue against your preferred option:

• "Steel-man the case for B as if I'm advocating it"
• "List 3 reasons A might fail that I haven't considered"
• "Who would prefer C and why?"

Document conclusions. Often surfaces hidden assumptions.

Time-boxed exploration

Explore-alternatives can spiral. Set a budget:

• "I'll spend ≤30 min generating + scoring options"
• "If no clear winner emerges, I'll defer to (e.g., 'pick the reversible option')"

Document the budget at start. If you exceed it: log "exceeded budget by X; outcome was Y" — learn for next time.

Output contract template

Save exploration to .supervibe/artifacts/specs/YYYY-MM-DD-<topic>-alternatives.md (or as section in brainstorm/PRD decision section/PRD).

Required sections:

1. Problem restated (1 paragraph)
2. Carbon-copy lookup (results from project-memory + code-search + research)
3. Options generated (≥3, each with 1-paragraph description)
4. Decision matrix (weights set first, scores second)
5. Sensitivity analysis (≥2 perturbations)
6. Adversarial scoring (≥1 steel-man for runner-up)
7. Recommendation (with rationale + acknowledged risks)
8. Confidence: high / medium / low
9. Decision-reversibility: reversible (low risk) / hard-to-reverse (high risk)

Anti-patterns

• Score-then-weight → post-hoc rationalization; weights bias toward preferred option
• Three options when one is straw-man → fake alternatives ("do nothing" priced ridiculously)
• No sensitivity analysis → brittle matrix invisible
• No adversarial scoring → confirmation bias unchecked
• No carbon-copy lookup → reinventing wheels; missed prior decisions
• Unbounded exploration → analysis paralysis
• Missing "do nothing" option → can't compare against status quo

Common workflows

Workflow: Library / vendor choice

1. Carbon-copy: any prior memory entries?
2. Generate 3-5 candidates (research via best-practices-researcher)
3. Dimensions: maturity, license, ecosystem, perf, cost, lock-in
4. Score; sensitivity; steel-man runner-up
5. Recommendation with reversibility note

Workflow: Architectural pattern choice

1. Carbon-copy: prior PRD decision sections in this area?
2. Generate 3 patterns (event-driven / sync RPC / async queue, etc.)
3. Dimensions: latency, complexity, ops cost, scaling envelope, blast radius
4. Score; heavy sensitivity (architecture is hard to reverse)
5. Steel-man preferred runner-up
6. Recommendation; flag reversibility = LOW

Workflow: Quick UX pattern choice (low stakes)

1. Carbon-copy: design system has it?
2. Generate 2-3 options
3. Dimensions: clarity, consistency, accessibility, effort
4. Score; light sensitivity
5. Recommendation; reversibility = HIGH

Verification

• Output contains all 9 required sections
• ≥3 options including "do nothing" / "keep current"
• ≥4 weighted dimensions
• Weights documented BEFORE scores
• Sensitivity analysis present
• Adversarial scoring present
• Confidence rubric: requirements; score ≥ 9

Related

• supervibe:brainstorming — uses this skill as core for option exploration
• supervibe:prd — uses this skill for the alternatives matrix
• supervibe:project-memory — for carbon-copy lookup
• supervibe:code-search — for code-pattern carbon copies
• supervibe:_ops:best-practices-researcher — for industry carbon copies
• supervibe:_ops:competitive-design-researcher — for design carbon copies