Community Evaluation

Drupal AI Skills Landscape

A data-driven look at what exists, what's missing, and where to focus

Date March 27, 2026 — Format Landscape analysis for the skills standardization effort
Context Drupal AI Initiative — community-driven quality standards for AI agent contributions

Key Findings

Section 1

The Context

The Drupal community is building a tracking system to standardize AI agent skills for contribution quality. Subject matter experts have identified the categories that matter most: UX, accessibility, security, test coverage, and coding standards. These are the things you'd tell a new contributor if they wanted their code to meet community expectations.

This evaluation maps the current public skill landscape against those goals. It covers every known Drupal-focused Claude Code skill repository, examines what each contributes, and identifies where the ecosystem has depth, fragmentation, or gaps.

All skills referenced here follow the agentskills.io specification — the open format now supported by 30+ AI coding agents including Claude Code, GitHub Copilot, Codex, Gemini CLI, Cursor, and others.


Section 2

Landscape Inventory

Nine known sources of Drupal-specific AI skills, ordered by combined GitHub stars:

Repository Stars Skills Primary Domains
scottfalconer (4 repos) 69 4 Contribution workflow, UX intent testing, issue queue triage, agent memory
madsnorgaard/drupal-agent-resources 35 5 Module dev, security review, D7-to-D10 migration, DDEV, code reviewer agent (plus 5 commands)
grasmash/drupal-claude-skills 29 9 Config management, contrib lifecycle, DDEV, OWASP security, OAuth, Search API, Acquia
jamieaa64/Drupal-DDEV-Setup 11 1 Full Drupal project lifecycle setup (new, onboard, sync, restore)
Omedia/drupal-skill 9 3 Frontend (Twig/CSS/JS), backend (modules/routing), tooling (DDEV/Drush)
ablerz/claude-skill-drupal-module 4 1 Module dev with 13 bundled API references, PHP 8.4+, auto-updates from drupal.org
nonzod/drupaldev-claude-skill 3 1 Coding standards enforcement (PHP, JS, CSS, Twig, YAML)
zivtech/drupal-meta-skills 0 7 Architecture planning (5 sub-planners), code review, config generation
zivtech/a11y-meta-skills 0 2 Accessibility planner-critic pair: WCAG implementation planning + ARIA/focus/semantic review
Also on drupal.org

The ai_skills module provides a library of Drupal-specific agent skills validated on every merge request. Issue #3565489 proposes first-class agentskills.io support in the Drupal AI module itself.


Section 3

Coverage Against Community Goals

The community identified six quality categories for contribution standards. Here is who covers each, and how deeply:

Category Depth Who Covers It Gap Analysis
Better UX Deep Scott Falconer (drupal-intent-testing) Only browser-driven UX verification skill for Drupal. Extracts intent from drupal.org issues, generates test manifests, produces evidence-based verdicts. No alternatives.
Accessibility Single source a11y-meta-skills (zivtech) One planner-critic pair exists (9-phase WCAG planning, ARIA/focus/semantic review), released as GPL-3.0. No alternative implementations yet — additional contributors would strengthen this category.
Security Moderate madsnorgaard (drupal-security), grasmash (OWASP patterns), ablerz (security API ref) Three options covering overlapping ground. No standard recommendation. The community would benefit from consolidation or a clear comparison.
Test Coverage Partial Scott Falconer (intent testing for E2E), madsnorgaard (general reviewer) Scott's intent testing covers UX-level E2E verification. No public skill addresses test strategy planning, coverage gap analysis, or test suite quality review.
Coding Standards Moderate nonzod (multi-language enforcer), ablerz (API refs + standards), grasmash (config standards) Multiple options, none dominant. ablerz is notable for auto-updating API references from drupal.org monthly via GitHub Actions.
Contribution Workflow Deep Scott Falconer (drupal-contribute-fix + drupal-issue-queue) Scott provides the strongest end-to-end contribution-quality lane, but not the only workflow-adjacent coverage. Kanopi covers issue formatting and merge-request flow, and grasmash covers contrib module management. The gap is depth and integration, not total absence.
Reading the depth ratings

Deep = dedicated skill with a structured protocol, unique in the ecosystem. Moderate = multiple options exist but fragmented; no standard pick. Single source = dedicated skill exists from one provider; no alternatives yet. Partial = covered tangentially but no dedicated skill. None = zero public skills address this category.


Section 4

Notable Skills

Five skills stand out for depth, uniqueness, or approach. Each solves a problem no other skill in the ecosystem addresses.

drupal-contribute-fix scottfalconer · 18 stars · GPL-2.0 Unique

The strongest end-to-end contribution-quality gate in the current landscape. It goes beyond workflow guidance by checking for existing upstream work, validating code quality, and packaging contribution artifacts before a maintainer sees them.

  • Searches drupal.org first to prevent duplicate contributions
  • Runs PHP lint + PHPCS on every patch before packaging
  • Never auto-posts to drupal.org — generates local artifacts for human review
  • Structured exit codes: redirect to existing fix (10), fixed in newer version (20), security issue (50)
  • Generates REPORT.md, ISSUE_COMMENT.md, and WORKFLOW.md as contribution artifacts
drupal-intent-testing scottfalconer · 5 stars · MIT Unique

Browser-driven, intent-focused UX verification. Rather than testing implementation details, it tests whether the user's original intent (from a drupal.org issue) is actually satisfied.

  • Extracts user intent from drupal.org issues, generates YAML test manifests
  • Three modes: guided watching, A/B paired comparison, exploratory fuzz testing
  • Uses semantic locators and accessibility tree snapshots — no vision model required
  • Safety-first: --safety read-only flag for fuzz mode; warns about destructive actions
  • Complements PHPUnit (unit/kernel) with real browser-level UX verification
a11y-planner + a11y-critic zivtech · GPL-3.0

The only accessibility skills in the Drupal AI skill ecosystem. A planner-critic pair: the planner designs accessible implementations before code is written, and the critic reviews them afterward.

  • a11y-planner: 9-phase WCAG implementation planning. Maps interactive patterns to WAI-ARIA APG, plans focus management, state communication, and visual accessibility
  • a11y-critic: ARIA pattern review, focus management audit, semantic validation, state communication checks
  • Open source, standalone package — additional contributors welcome
claude-skill-drupal-module ablerz · 4 stars · GPL-2.0

A different approach to keeping AI agents current: rather than encoding knowledge in prompts, this skill bundles 13 API reference documents and auto-updates them from drupal.org monthly via GitHub Actions.

  • Covers hooks, entities, forms, plugins, routing, security, testing, cache, access control, DI patterns
  • Monthly CI workflow checks for breaking changes from official drupal.org sources
  • Generates complete installable modules with PHP 8.4+ attributes (not annotations)
  • Most sustainable approach to knowledge currency in the ecosystem
drupal-issue-queue scottfalconer · 6 stars · Python Unique

Read-only skill for searching and triaging the Drupal.org issue queue. Prevents wasted effort by surfacing existing work before contributors start from scratch.

  • Three backends: drupal.org API, drupalorg-cli, or auto-detect
  • JSON and Markdown output formats for flexible integration
  • Hands off to drupalorg-cli for fork/merge request actions
  • Pairs with drupal-contribute-fix to form a complete contribution pipeline: find (issue-queue) → verify (intent-testing) → package (contribute-fix)

Section 5

Recommendations

Based on coverage analysis, uniqueness, and alignment with the community's stated quality categories:

Tier 1 — Should Be Standard

Unique skills with no alternatives. High impact on contribution quality.

  • drupal-contribute-fix (scottfalconer) — the strongest contribution quality gate. It addresses the asymmetric pressure problem more completely than the workflow-helper alternatives.
  • drupal-intent-testing (scottfalconer) — the only browser-based UX verification for Drupal. Fills the gap between "tests pass" and "it actually works for users."
  • a11y-planner + a11y-critic (zivtech) — currently the only accessibility skills in the ecosystem. Accessibility cannot be optional in contribution standards; more implementations are needed.

Tier 2 — Important, Community Should Pick One

Multiple options exist. The tracking system should recommend a standard choice.

  • Security: madsnorgaard/drupal-security, grasmash OWASP patterns, or ablerz security refs. All cover OWASP Top 10 differently. Consolidation or a clear comparison would help.
  • Coding standards: ablerz's auto-updating approach is the most sustainable (monthly CI from drupal.org), but nonzod covers more languages. Pick based on maintenance model preference.
  • DDEV/setup: jamieaa64 has the most complete lifecycle (new project, onboarding, sync, restore). grasmash and Omedia cover similar ground with different scopes.

Tier 3 — Valuable for Specific Use Cases

Deep skills for contributors working in specialized areas.

  • Drupal architecture: drupal-meta-skills (zivtech) bundles a planner stack, code review, and config generation into a single GPL-3.0 package.
  • Search API: grasmash/drupal-search-api for Solr/Elasticsearch configuration.
  • Agent memory: Scott Falconer's self-learning-skills (40 stars) for cross-session knowledge persistence. Infrastructure, not a quality gate, but useful for agents that work on Drupal projects long-term.

Section 6

The Gap That Matters Most

The biggest problem facing contributors isn't a missing skill. Forty-plus skills exist across nine-plus repositories, each with different installation methods, documentation styles, and quality levels. A contributor who wants to "write code with proper security" has to discover that three different skills exist, compare them, and make a choice with no guidance.

The tracking system the community is building solves exactly this. The data in this evaluation shows clear winners in underserved categories and clear fragmentation in others. What's needed is a curated, maintained index that maps skills to quality categories and signals which are recommended, which are alternatives, and which cover niche use cases.

The agentskills.io specification gives us a shared format. The proposal for first-class skill support in the Drupal AI module gives us a distribution mechanism. The community's SME-driven quality categories give us the organizing framework. All that's left is connecting the pieces.