Lamina decision making
Headless Product Design skill for AI coding agents | Design how it works, verify what you ship.
npx -y skills add aryaniyaps/lamina --skill lamina-decision-makingAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Prioritize contract gaps and verify findings — impact × effort, primary actor filter. Use when reconciling conflicts or ranking findings[].
SKILL.md
2.5 KB, as published. Nobody here has run it
Decision Making (agent-native)
Rank contract changes and verify findings by impact on primary actor goals and invariants — output to decisions.md, sorted findings[], and fix.md sections.
Fix target classification
Before writing fix.md, set fix_target on each finding in run.json:
| Signal | fix_target |
|---|---|
| Missing UI, bug, invariant violation on live product, a11y on existing screen | product |
| Scope change, new workflow/invariant, defer feature, contract wrong | contract |
Default: product when fix_target is unset. Product fixes belong in fix.md → coding session. Contract items belong in fix.md → /lamina-design delta prompt only.
Decision filter
Does this finding change a specific design or implementation decision? If not, dig deeper or drop.
Impact × effort (verify findings)
| Impact | high | medium | low |
|---|---|---|---|
| Definition | Blocks workflow, invariant violation, primary actor stuck | Friction with workaround | Polish |
| Effort | high | medium | low |
|---|---|---|---|
| Definition | Cross-workflow contract change | Localized screen/scenario | Copy/label fix |
Sort: high impact + low effort first; then high impact + high effort. Record rationale in decisions.md.
Primary actor filter
When actors conflict:
- Does option serve primary persona goals completely?
- If no → deprioritize or cut; if yes → design fully for their scenario.
- Log trade-off in
decisions.mdwith rejected actor impact noted.
Used when reconciling parallel persona panel outputs.
Evidence weighting
| Source | Weight |
|---|---|
| Invariant probe failure on live product | highest |
| Actor walk blocker (repro steps) | high |
| Walkthrough/repo grounding | medium |
| Assumption in design | low — mark to-verify |
Never rank invented analytics above reproduced actor failures.