validate-work

Verify a work candidate against acceptance criteria and report evidence-backed results.

1|Updated Apr 15, 2026
One-click install
npx skills add https://github.com/pnewsam/skills --skill validate-work-pnewsam
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: validate-work
Source: https://github.com/pnewsam/skills/tree/main/registry/validate-work
Command: npx skills add https://github.com/pnewsam/skills --skill validate-work-pnewsam

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? It answers whether a change actually meets its acceptance criteria by gathering real evidence instead of assuming tests or passing CI prove correctness, while clearly labeling what remains unverified. ## Core Features & Use Cases - Scoped validation: Run a targeted check, selected criteria, or a full assessment, with partial results explicitly labeling unchecked requirements. - Surface-aware proof selection: Map each changed surface (source, tests, config, lockfiles, schemas, infrastructure, docs, UI) to its strongest practical evidence using the verification reference. - Visual and UI evidence helpers: Check WCAG contrast ratios, lint spacing values against a scale, capture static HTML renders with headless Chrome, and diff before/after screenshots deterministically. - Honest reporting: Distinguish reproduced baseline failures, introduced regressions, flakes, and unavailable evidence; never edit code, weaken thresholds, or convert missing checks into passes. - Use Case: After implementing a UI change, validate it by running the repo's configured checks, capturing before/after screenshots with shot_diff.mjs, verifying contrast with check_contrast.py, and producing a per-criterion pass/fail/unverified report. ## Quick Start Use the validate-work skill to verify my current changes against the acceptance criteria in the issue and report which requirements pass, fail, or remain unverified.

Frequently Asked Questions about validate-work

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate a code change against acceptance criteria?▼

Map each acceptance criterion to its strongest practical evidence, run the repository's configured checks and relevant regression boundaries, then report pass, fail, or unverified per criterion. Partial validation must label unchecked requirements rather than presenting them as complete.

How do I check WCAG contrast ratios from the command line?▼

Run check_contrast.py with a foreground and background color in hex or RGB form to get the contrast ratio and AA/AAA verdicts. It also accepts a JSON file of color pairs and exits nonzero if any pair fails WCAG AA.

How do I compare two screenshots to detect visual regressions?▼

Capture before and after PNGs with any headless renderer, then run shot_diff.mjs on the pair. It decodes the PNGs with no third-party dependencies, reports the changed pixel ratio, and can write a highlighted diff image and JSON report.

Can validation reuse previous test or CI results?▼

Prior evidence can be reused only when its candidate and relevant assumptions remain valid. Changes to configuration, dependencies, base, runtime, or requirements can invalidate earlier results even when source files did not change.

What are the limitations of static HTML screenshot capture?▼

The render helper captures supplied static HTML with local headless Chrome and never installs a browser. It cannot prove a real app's behavior involving authentication, external assets, or interactive states; those require the application's actual browser harness.

Does validation fix failing code or publish results?▼

No. Validation only gathers and reports evidence; it never edits implementation or tests, weakens thresholds, commits, publishes, or merges. Repairs return to execution, and a separate report is written only when requested.