skills/local/stelow-product-interface-alternatives/SKILL.md
[stelow] Interface alternatives exploration skill. Use when generating interface proposals using the 5-archetype library. Produces 1, 3, or 5 proposals depending on appetite, plus hybrid recommendation for Core/Complete. Part of stelow but can be used standalone.
npx skillsauth add renatocaliari/agent-sync-public-skills stelow-product-interface-alternativesInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Tools: See
references/cli-tools/subagents.mdfor subagent patterns.
This skill executes the Interface Alternatives phase.
The orchestrator reads this file directly when needed.
This skill works standalone. Use the Input Detection section below to tell the skill what interface you want to brainstorm. Follow the instructions inline.
Standalone awareness: when inside stelow, reads appetite from .stelow/*/spec-product.md. When standalone, defaults to Core appetite (3 interfaces + hybrid). If the spec-product.md is not in the expected path, pass it explicitly or let the LLM scan for spec-product*.md in the current directory.
Step 0: Read appetite from spec-product.md and choose interface exploration depth.
Appetite controls how many interface alternatives are explored. Quality is not cut; the number of explored alternatives is.
| Appetite | Interface exploration |
|----------|----------------------|
| Lean | 1 suggested interface only. No alternative exploration. |
| Core | 3 archetypes explored + 1 hybrid recommendation. |
| Complete | 5 archetypes explored + 1 hybrid recommendation. |
WF_DIR="$(ls -td .stelow/*/*/ 2>/dev/null | head -1)"
# shellcheck disable=SC1091
source "$(dirname "${BASH_SOURCE[0]:-$0}")/../../stelow-product-orchestrator/references/cli-tools/read-config.sh"
APPETITE=$(stelow_read_appetite)
# Fallback to spec-product.md frontmatter if no stelow/index
if [ -z "$APPETITE" ]; then
APPETITE=$(grep -oP '^appetite:\s*\K\S+' $(ls -t .stelow/*/*/plans/spec-product*.md 2>/dev/null | head -1) 2>/dev/null || echo "Core")
fi
case "$APPETITE" in
Lean)
INTERFACE_COUNT=1
ARCHETYPES="A" # Choose the archetype that best fits the work pattern.
HYBRID="skip"
;;
Core)
INTERFACE_COUNT=3
ARCHETYPES="A,D,E" # Safe baseline + simplicity + expert flow; replace one with B/C if better justified.
HYBRID="yes"
;;
Complete)
INTERFACE_COUNT=5
ARCHETYPES="A,B,C,D,E"
HYBRID="yes"
;;
*)
echo "APPETITE_UNKNOWN: '$APPETITE'. Defaulting to Core interface exploration."
INTERFACE_COUNT=3
ARCHETYPES="A,D,E"
HYBRID="yes"
;;
esac
Step 1: Read the references/ files to guide the process:
| File | Covers | When to read |
|---|---|---|
| references/interface-context.md | Progressive Clarification, when to use, system equivalents | Before starting |
| references/interface-reconstruction.md | Context reconstruction, hidden job extraction | Before generating |
| references/interface-rules.md | Separation Rule, Work-Pattern-First Composition, Smell Self-Audit, State Coverage Baseline, Forced Trade-Off Rule, output quality | Before generating |
| references/archetypes.md | 5 archetypes with descriptions | During generation |
| references/hybrid-recommendation.md | Hybrid recommendation strategy | Step 3 only |
Use the subagents tool (see references/cli-tools/subagents.md) to generate the appetite-selected proposals in parallel. For Lean, run one worker only. For Core, run 3 workers. For Complete, run 5 workers.
$INTERFACE_COUNT parallel workers (fresh context, explicit reads):
Generate only the selected archetype(s) from $ARCHETYPES
⚠️ CRITICAL — Before generating, each worker MUST read:
1. references/interface-rules.md — Work-Pattern-First Composition (mandatory Section 0)
2. references/interface-rules.md — Smell Self-Audit (mandatory Section 6)
3. references/interface-rules.md — State Coverage Baseline (mandatory Section 7)
4. references/output-format.md — full output format with all 8 sections
5. spec-product.md — body + frontmatter (`appetite`, `review_mode`, `domains_detected`)
6. tech-recon.md — tech constraints (if it exists)
Each worker MUST receive `reads: [spec-product.md, tech-recon.md]` (or `references/interface-*.md` when running standalone).
Each worker runs `context: "fresh"` — proposals must be independent of orchestrator deliberation history.
Each outputs to .stelow/{date}/{dir}/interfaces/proposal-{letter}.md
.stelow/{YYYY-MM-DD}/{_dir}/interfaces/interfaces_{v}.mdStep 2: see references/output-format.md for instructions to format and concatenate selected proposals.
CRITICAL: Hybrid is generated only when $HYBRID = yes (Core or Complete appetite) and AFTER all selected proposals are complete to avoid bias.
Use the subagents tool (see references/cli-tools/subagents.md) to merge:
Agent: worker
Task: Generate Hybrid Proposal
Reads: selected proposal files
Output: Append to interfaces.md per hybrid-recommendation.md
## Visual Review (Interface Gate — Automatic)
**After all selected proposals (+ Hybrid when applicable), use the Plannotator gate** (see `references/cli-tools/plannotator.md` for the correct command). Execute it directly — do NOT describe it to the user.
Prefer the `plannotator` tool (registered by stelow extension). Fall back to bash CLI if unavailable:
plannotator filePath=.stelow/{YYYY-MM-DD}/{_dir}/interfaces/interfaces_v{N}.md
plannotator annotate .stelow/{YYYY-MM-DD}/{_dir}/interfaces/interfaces_v{N}.md --gate --json
Wait for the decision. If `approved`, the tool auto-creates the receipt. Then advance to Interface Selection.
Then use **Pattern 2** from `references/cli-tools/ask.md` to let the user pick one proposal. Do NOT just describe what comes next — execute it.
## User Selection (Interface Selection)
After visual review and approval, use **Pattern 2** from `references/cli-tools/ask.md` to ask the user which proposal to follow.
### Save Selected Interface as Permanent Artifact
After the user picks a proposal, extract it from `interfaces_{v}.md` into a dedicated artifact:
```bash
PROPOSAL_LABEL="{chosen_proposal_label}" # e.g. "A — Convention Navigator"
OUTPUT=".stelow/{YYYY-MM-DD}/{_dir}/interfaces/selected-interface.md"
# Extract the chosen proposal's full content (all 7 sections)
awk -v label="$PROPOSAL_LABEL" '
/^## Proposal / && !started { started = ($0 ~ label); next }
/^## Proposal / && started { exit }
started { print }
' ".stelow/{YYYY-MM-DD}/{_dir}/interfaces/interfaces_{v}.md" > "$OUTPUT"
# Verify: chosen interface must contain Work Pattern and Breadboarding
grep -q "Work Pattern" "$OUTPUT" || { echo "ERROR: selected interface missing Work Pattern Declaration"; exit 1; }
grep -q "Breadboarding" "$OUTPUT" || { echo "ERROR: selected interface missing Breadboarding section"; exit 1; }
The saved artifact contains the complete proposal: work pattern declaration, breadboarding guidelines, ASCII wireframe, interaction flow, trade-off analysis, design smell audit, and state coverage table.
The tech planning and execution stages reference this file for UI direction — they should NOT regenerate or reinterpret the interface.
Interface proposals are saved to:
.stelow/{YYYY-MM-DD}/{_dir}/interfaces/interfaces_{v}.md
The chosen interface (after user selection) is saved to:
.stelow/{YYYY-MM-DD}/{_dir}/interfaces/selected-interface.md
When called outside the workflow with no pre-existing spec-product.md:
Input:
├── User provided a spec-product*.md path?
│ └→ Read its IN/OUT scope and solution description
├── User described the feature verbally?
│ └→ Extract: what does the user need to accomplish?
└── No structured input?
└→ Ask: "What interface do you want to explore? Describe
what the user needs to do, and any constraints (platform,
existing UI patterns, brand guidelines)."
Then follow the appetite-selected archetype generation process above.
If a tool is unavailable, check:
references/cli-tools/
tools
Extrai métricas estruturadas, cálculos e estimativas de transcripts de entrevistas com clientes do Sommelier de IA. Produz um JSON com dores, frequências, tempo gasto, pessoas envolvidas, economia potencial, ROI e recomendações financeiras. Projetado para alimentar o cali-degustia-diagnostico ou integrar com dashboards/planilhas.
tools
Guia a coleta de depoimentos de clientes do Sommelier de IA no momento certo do processo, usando a abordagem de Hormozi: pedir depois da primeira evidência de resultado, nunca na entrega. Gera depoimentos mais autênticos e reduz a sensação de que o cliente está sendo "solicitado".
development
[stelow] Full UX critique for visual interfaces. Accepts a live URL, source code directory, or screenshot image. Evaluates accessibility (WCAG AA), Nielsen's 10 heuristics, visual hierarchy, cognitive load, consistency, mobile responsiveness, AI slop, emotional journey, and design personas — then generates a classified gap report. Standalone or integrated into stelow and stelow-product-testing-execution.
development
Building trust through perception and guarantee mechanisms. Covers ten pillars to materialize trust, guarantee types from unconditional to anti-guarantees, and strategic approaches for different contexts.