Skip to content
Skills Library
Design & UX · Testing

forma-validation

/forma-validation

Validate tasks, accessibility, rendering, states, and fidelity with proportionate evidence and reproducible findings. Use for QA and confirmation after changes; not to certify inaccessible areas or treat isolated builds/tests as visual proof.

Repository
forma ↗
Licence
MIT
Last repository change
11.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 11.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Validate tasks, accessibility, rendering, states, and fidelity with proportionate evidence and reproducible findings. Use for QA and confirmation after changes; not to certify inaccessible areas or treat isolated builds/tests as visual proof.

Original author description

Validate tasks, accessibility, rendering, states, and fidelity with proportionate evidence and reproducible findings. Use for QA and confirmation after changes; not to certify inaccessible areas or treat isolated builds/tests as visual proof.

How it works

  1. When to use
  2. Inputs
  3. Procedure
  4. Output and stopping condition
  5. Coordination and execution boundaries
Original SKILL.md headings · 50 lines, 680 words
SKILL.md preview
---
name: forma-validation
description: "Validate tasks, accessibility, rendering, states, and fidelity with proportionate evidence and reproducible findings. Use for QA and confirmation after changes; not to certify inaccessible areas or treat isolated builds/tests as visual proof."
metadata:
  version: 1.36.0
---

# Validation and quality

## When to use

- Use for audits, acceptance, regression, accessibility, rendered comparison, and confirmation of integrated capabilities.
- Do not invent scores, declare full compliance from automation, or expand an audit into unauthorized edits.

This module must be read by the agent; it does not trigger tests, fixes, or other modules by itself.
Preserve scope, brand, and stack. Distinguish verified, inferred, untested, and not applicable, with a reason.

For a relevant saved capability, use [reuse by capability](../forma/references/capability-reuse.md): search by the task, not the source project, and bind the mechanism to the destination's own identity and state owners. Read only the matching contract; do not expand a local fix into a redesign. Source names and exact profiles remain provenance or explicit fidelity targets.

## Inputs

- Artifact/route, revision actually loaded, task, acceptance criteria, and supported states/devices.
- A baseline for comparisons, data/fixtures, and tools/environments that are actually available.

## Procedure

1. Build a matrix proportionate to the risk and request before judging the result.
   Read [verification](../forma/references/verification.md); select dimensions and minimum proof without requiring every dimension for a small fix.
   When consuming agent/renderer gates, check [coverage and receipts](../forma/references/verification.md#verifier-coverage-and-receipts): a pass without measurements, zero assessed items, or an old revision does not approve the current delivery.
2. Confirm execution identity and follow the task through input, action, success, and recovery.
   Use [rendered validation](../forma/references/rendered-validation.md); verify persistence after reopening if promised, and distinguish an in-memory prototype from real operation.
3. For access, read [accessibility](../forma/references/accessibility.md) and exercise the relevant keyboard, focus, semantics, zoom, and alternative paths.
   Inspect computed accessible names for inputs presented with icons or placeholders; descriptions explain a field but do not replace its label.
   If layout/color is affected, use [spatial contracts](../forma/references/layout-spatial-contracts.md) and [color refinement](../forma/references/color-refinement.md).
   For a new surface or broad visual refinement, inspect the [coordinated foundation](../forma/references/verification.md#coordinated-foundation-review) in representative consumers and supported modes. Isolated token or contrast checks do not establish a coherent composition.
4. For reference work or reuse, test the operation that distinguishes the requested mechanism from an approximation.
   Consult reuse in [playbooks](../forma/references/playbooks.md) and [fidelity](../forma/references/reference-fidelity.md); observe the primary path, fallback, and relevant states separately. The [component lab](../forma/references/reusable-components.md#run-the-lab) exposes actual renderer and state diagnostics; counters and geometry tests still need rendered inspection, Save/reload, error recovery, unmount/remount and BFCache checks in the consumer.
   When frozen reference values exist, run the [fidelity verifier](../forma/references/fidelity-contracts.md) against consumer measurements. Missing fields, NaN, omitted points, or different revisions cannot pass through averaging. A passing numerical fixture does not prove that the browser loaded the revision or that the contract covers everything; confirm those facts separately.
5. Record failures with reproduction steps, effects, severity, and the smallest correction; fix only within existing authorization.
   Recheck affected states and critical paths, preserving before/after captures. Additional tests must address a remaining concern.
6. When changing Forma itself, use [plugin operation](../forma/references/plugin-operation.md) and the [evaluation protocol](../forma/evals/routing-scenarios.md#compare-versions-without-confusing-evidence).
   Package integrity, host discovery, agent decisions, and consumer operation have separate acceptance criteria. Apply [release validation](../forma/references/release-validation.md) when the request concerns a revision, independent trials, reusable components, or observed retrieval: freeze inputs/submissions, keep post-review corrections separate, and record each browser/input/motion/WebGL/cost gate as passed, failed, unverified, or inapplicable with a reason. Two Chromium shells are not two engines; narrow width is not touch, and CPU render time is not GPU cost.

## Output and stopping condition

Deliver results by criterion, evidence/environment, actionable findings, and coverage limits.
Helper scripts verify their input contracts; passing them does not replace observation in the product.
Stop when relevant criteria have evidence and no material blocker remains, or conclude the audit with clearly open failures.
Without runtime access, state what static analysis confirms and what remains untested; do not treat a limitation as approval.

## Coordination and execution boundaries

Consult the relevant Forma domain specialist to interpret evidence. Browser, accessibility and end-to-end checks remain this module's responsibility, using the available runner and existing project tests.
The coordinator selects additional modules; do not recursively dispatch between specialists. Use the available tools according to [capabilities](../forma/references/integrations.md); a tool name does not prove connection or execution.
Do not create suites that merely mirror the implementation. Test evidence, visual inspection, and task validation answer different questions.
Files in this package (2)
  • SKILL.md 6 KB
  • agents/openai.yaml 1 KB

Versions

RepositoryScoreLicence
This page · figueirego/forma100MIT
figueirego/forma100MIT

Database snapshot: 13.11.2026 · package version bffffea89293 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 90forma-product
    Refine ambiguous design requests into shared briefs and executable prompts, investigate user tasks, and design product flows with proportionate research. Use for unresolved intent, sequence, or understanding; not for isolated color adjustments or fabricated user evidence.
  • 90forma-product
    Refine ambiguous design requests into shared briefs and executable prompts, investigate user tasks, and design product flows with proportionate research. Use for unresolved intent, sequence, or understanding; not for isolated color adjustments or fabricated user evidence.
  • 100forma
    Set up and use Forma, a headless CLI that renders designs from JSON frames and shared design tokens to PNG. Use this skill whenever the user mentions forma, a repo contains forma/config.json, or the user wants to add a design system (colors, type, logos, example images) to a…
  • 100forma
    Read and edit Forma boards, the designer's research boards that live as plain folders (board.json + media/ + components/). Use whenever the user asks to add, sort, connect, find or change anything on a Forma board, to bring research in (images, videos, links, notes, meetings,…
  • 100forma-templating
    Builds an installable project template (a separate plugin in the forma marketplace's templates/ directory) out of the current Forma project, under the name the human gives. Invoke when the human says "create a project template", "make a template from this project", "package the…
  • 100trove-design-review
    Design review skill for evaluating UI/UX implementations against design specs. Checks visual fidelity, interaction patterns, and design system compliance. Use when: design review; review the design; visual fidelity check.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.