Skip to content

adityaarsharma/orbit

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

114 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🪐 Orbit

Complete UAT for WordPress Plugins — Now Agentic

A Claude Code plugin · 116 runtime-evergreen /orbit-* skills · 10-agent QA team · CTO → PM → Dev → QA → Security → Release

v3.0 — Orbit Agentic. Orbit is no longer just a skill suite. It's a 10-agent QA team connected to a shared brain (brain-posimyth). CTO's brain is the team's constitution — every agent reads it first. Approved patterns get remembered. Cold starts become warm starts.

The one-command audit:

/orbit-do-it ~/plugins/my-plugin

Auto-detects plugin type. Picks the right pipeline. Runs core audits + UAT + perf + security + compat in parallel. Writes a one-page TL;DR + a master HTML report. Walks away. Comes back to a verdict.


PHP WordPress Playwright Stagehand Lighthouse Claude Code Agentic


👨‍💻 Dev · zero-regression releases  ·  🧪 QA · structured coverage + auto-generated specs  ·  📊 PM · flow maps + RICE backlog + release notes  ·  🎨 Designer · visual diffs + token audits + dark mode  ·  🚀 Release Ops · WP.org gates + EU CRA compliance  ·  👤 End User · real browser, real flows, AI-resolved tests

📖 VISION.md  ·  🚀 Skills Reference  ·  🌱 Runtime-Evergreen Pattern  ·  🛡️ Evergreen Security  ·  🤖 Orbit Agentic  ·  📓 Changelog

Install in 60s · Orbit Agentic — v3.0 · The brainless agent · The 116 skills · Runtime-evergreen, explained · Role guide · GitHub


What Orbit Is

A Claude Code plugin that gives a WordPress plugin team — dev, QA, PM, designer, release ops — a single command (/orbit-do-it) that audits everything that matters before a release: code standards, security, performance, accessibility, UAT, visual regression, hosting compatibility, EU CRA compliance, and 100+ other concerns.

It's not a SaaS. Runs locally via Docker (wp-env) + Claude Code. No accounts, no subscriptions, no cloud. The whole stack — 116 skills, all the scripts, the installer, this README — lives in one Git repo.

It's runtime-evergreen. When a skill runs, it fetches the canonical source-of-truth doc (e.g. Elementor's changelog, NVD's CVE feed, Kinsta's banned-plugins page) and applies today's rules — not a snapshot from when the skill was written. The same /orbit-elementor-compat SKILL.md handles V4 today, V5 next year, V6 the year after. Without anyone editing it.

It composes with WordPress/agent-skills — WP core's official AI agent skills (Brandon Payton, January 2026). Orbit's installer chains npx openskills install WordPress/agent-skills, so users get both: WP core's runtime/Playground primitives + Orbit's QA/UAT/audit suite.


How it works — 3 layers

YOU
 │  "UAT audit NexterWP v2.5"          ← natural language in Claude Code
 ▼
AGENT  (05-uat.md)
 │  Step 1: Brain Prime                 ← 5 searches on brain-posimyth
 │    "What did the last UAT find?"
 │    "Are there known flaky tests?"
 │    "What WP standards apply here?"
 │  Step 2: Spin Docker WP env
 │  Step 3: Playwright E2E
 │  Step 4: Dispatch 07-Security + 06-Perf + 04-Designer in parallel
 │  Step 5: Severity triage → CLEAR or BLOCKED
 │  Step 6: Ingest findings to brain   ← [uat, bug, nexterwp, High, ...]
 ▼
SKILLS  invoked by the agent automatically
 │  /orbit-playwright      → runs E2E browser tests
 │  /orbit-visual-regression → screenshots, diffs
 │  /orbit-wp-security     → XSS/CSRF/SQLi scan (via 07-Security)
 │  /orbit-lighthouse      → Lighthouse score (via 06-Performance)
 ▼
MCP + TOOLS  that skills use
    brain-posimyth          ← read history, write findings
    wp-env (Docker)         ← clean WP install for testing
    Playwright + Chrome     ← real browser, real flows
    gh CLI                  ← open issues, create PRs

The brain is what makes it a team, not just a tool. Every finding is ingested. Every approved pattern is remembered. Every redline is surfaced the next time the same task runs. The agents get smarter every sprint — without you changing any files.


Orbit Agentic — v3.0

"Skills are easy. Process is harder. Brain is evergrowing — like onboarding a new person who's however smart, but still needs to learn YOUR products."

v3.0 turns Orbit into a 10-agent QA team where each agent has a defined role, written SOPs, a dedicated brain collection, and the MCP access to act on what they find. The more you use it, the smarter the whole team gets.


The 10-Agent Team

# Agent Role in one line
00 CTO Strategic advisor. Reads all 10 brains. Sets direction — never executes. Sole writer to the shared brain.
01 PM Daily coordinator. RICE scoring, feedback mining, sprint health. Routes every task to the right specialist.
02 Code Reviewer Senior + skeptical. PHP, Gutenberg, Elementor, compat. APPROVE / REQUEST CHANGES / NITPICK — with file:line.
03 Senior Dev Builds features, fixes UAT bugs. Runs WP standards before done. Never self-merges.
04 Dev Designer WCAG 2.2 AA, RTL, dark mode, empty/error states. Writes design specs — 03 implements them.
05 UAT Docker WP env, Playwright E2E, visual regression. Orchestrates 06 + 07 + 04 in parallel. Severity gates releases.
06 Performance Hook weight, DB queries, bundle analysis, Lighthouse. Sets perf budgets. Enforces regression thresholds.
07 Security XSS, SQLi, CSRF, supply chain, CVE, Stripe/EDD/Freemius, GDPR, PCI, premium gating. NEVER tests production.
08 Release 7-step gate, WP.org Plugin Check, zip hygiene, release notes (POSIMYTH voice), cross-channel announce.
09 Docs README, feature docs, hook reference, in-code comments, changelog language. Ships with release — never after.

The shared brain — CTO is the head

orbit/00-cto is the team's constitution. Every agent reads it first — before their own collection. The CTO agent is the only one that writes to it. What lives there:

brain-posimyth
└── orbit/
    ├── 00-cto/
    │   ├── hard-rules/       ← WP coding standards, security patterns, release rules
    │   ├── decisions/        ← Technology + product direction decisions
    │   ├── competitor-intel/ ← Competitor moves, market signals
    │   ├── risks/            ← Unstable APIs, CVE trends, deprecation warnings
    │   └── approved-patterns/← Patterns promoted from any agent to team-wide
    │
    ├── 01-pm/        ← Roadmap, RICE decisions, sprint history
    ├── 02-code-reviewer/ ← Review patterns, approvals, redlines
    ├── 03-senior-dev/    ← Build patterns, fix history
    ├── 04-dev-designer/  ← WCAG findings, RTL patterns, token decisions
    ├── 05-uat/           ← Bug reports, UAT results, flaky test registry
    ├── 06-performance/   ← Benchmarks, perf budgets, regression history
    ├── 07-security/      ← CVE findings, vuln patterns, payment audit history
    ├── 08-release/       ← Release history, WP.org rejections, announce templates
    └── 09-docs/          ← Freshness tracking, API doc history, voice patterns
# First install — seed 40 knowledge drawers into orbit/00-cto/hard-rules/
bash brain/seed-brain.sh --key <your-orbit-admin-key>

Day-one intelligence in the CTO brain: WP escaping rules, block.json required fields, WCAG 2.2 AA checklist, Stripe webhook security, readme.txt rejection patterns, N+1 DB query patterns, and 34 more. No cold starts for any agent.

Two keys:

  • Team key — read orbit/00-cto + own collection. Agents recall past findings, approved patterns, known issues.
  • Admin key — full read + write. Ingest findings, promote patterns, announce cross-channel. EDD ops: Admin only.

Brain Prime — what every agent does first

Before touching any code or producing any output, every agent runs 5 brain searches and writes a Brain Prime block:

BRAIN PRIME — NexterWP v2.5 (UAT)
• CTO rules:   Never ship unescaped output. RTL mandatory. Lighthouse target ≥ 85.
• Bug history: v2.4 block reorder crash (orbit/05-uat/nexterwp). Fixed in v2.4.1.
• Patterns that worked: Docker WP 6.8 + Gutenberg 18.x env. Playwright --project=chromium first.
• Patterns to avoid: waitForTimeout() — caused 3 flaky tests in v2.3 audit.
• Open question:  Is scroll-animation block new in v2.5? (brain silent — will check changelog)

This block is pinned before any skill invocation. The agent never re-asks for context that's already in brain.


The approval loop

Every approve and revise from the operator teaches the brain:

you: approve              → agent asks "Save as approved pattern?" → ingests to own collection
you: revise: <why>        → agent auto-ingests redline → surfaces this FIRST next time same task runs
you: skip                 → ingests as deprioritised — agent won't suggest it again

CTO promotes team-wide: When a pattern is strong enough for the whole team (not just one agent), Admin runs:

# Example: promote a new nonce pattern to team-wide hard rule
# Admin ingests to orbit/00-cto with [cto, hard-rule, ...] tag
# Every agent picks it up on next Brain Prime

How agents collaborate — 5 real scenarios

Scenario 1 — New feature, end-to-end

A feature request ("Add scroll animation block to NexterWP") flows through the whole team:

01-PM        → RICE score: Impact 8 / Confidence 7 / Effort 5 → score 112 → APPROVED
               Routes to: 03-SrDev (build) + 04-DevDesigner (spec first)

04-DevDesigner → Brain Prime: loads WCAG rules from orbit/00-cto, past RTL findings from orbit/04
               → DESIGN SPEC: RTL mirror required. Reduced motion variant required. Touch target ≥ 44px.
               → Routes spec to: 03-SrDev

03-SrDev      → Brain Prime: loads WP standards from orbit/00-cto, past build patterns from orbit/03
               → Builds. Runs /orbit-wp-standards before PR.
               → Handoff brief to: 02-CodeReviewer (via 01-PM)

02-CodeReviewer → Brain Prime: loads PHP hard rules, past TPA redlines from orbit/02
               → Reviews PHP + block.json + Gutenberg + compat
               → REQUEST CHANGES: "save() uses SSR — must declare RenderCallback in block.json"
               → 03-SrDev fixes → re-review → APPROVE

05-UAT        → Brain Prime: loads severity rules, v2.4 bug history from orbit/05
               → Playwright E2E + visual regression. Dispatches 07-Security + 06-Perf + 04-Designer in parallel.
               → All pass → UAT CLEAR. Routes to: 08-Release

08-Release    → 7-step gate. All pass.
               → Release notes drafted. Cross-channel announce.
               → Routes to: 09-Docs (same day publish)

09-Docs       → Freshness audit. Feature documented. API hook reference updated. Publish same day as release.

Scenario 2 — Critical security found mid-sprint

07-Security   → Scanning NexterWP v2.5 RC
               → CRITICAL: Settings page — /wp-admin/admin.php?page=nxtwp echoes ?search= without esc_html()
               → ESCALATING CRITICAL immediately to 01-PM. Stopping scan.
               → Ingests to orbit/07-security/nexterwp: [security, nexterwp, Critical, xss-settings-page, v2.5-rc]

01-PM         → Receives escalation. Blocks sprint. Routes to 03-SrDev as Priority 0.
               → Notifies 08-Release: release gate will not run until Critical is resolved.

03-SrDev      → Fixes: esc_html( sanitize_text_field( $_GET['search'] ) )
               → Routes back to 07-Security for re-scan.

07-Security   → Re-scans. Clean. Confirms fix.
               → Ingests: [security, nexterwp, fixed, xss-settings-page, v2.5]
               → Routes to 05-UAT for regression test.

05-UAT → 08-Release → 09-Docs  (normal flow resumes)

Scenario 3 — WP.org rejection: the brain learns forever

08-Release    → Submitted NexterWP v2.4.0 to WP.org.
               → REJECTED: "Plugin is loading scripts/styles on all admin pages"

08-Release    → Ingests rejection to orbit/08-release:
                 [release, nexterwp, wp-org-rejection, scripts-all-admin-pages, v2.4.0]
               → Routes to 00-CTO: "This may be a team-wide pattern issue"

00-CTO        → Checks orbit/02-code-reviewer — same pattern in TPA code too.
               → Decision: promote to hard rule.
               → Ingests to orbit/00-cto/hard-rules/:
                 [cto, hard-rule, no-scripts-all-admin-pages, wp-org-requirement, 2026-05-20]

From now on:  Every agent reads this rule on Brain Prime.
               02-CodeReviewer blocks any PR that loads scripts on all admin pages.
               08-Release checks for it in the 7-step gate.
               One rejection — zero repeats, across all 3 plugins, forever.

Scenario 4 — Performance regression caught before release

06-Performance → Benchmark NexterWP v2.5 vs v2.4 baseline (orbit/06-performance/nexterwp/budget)
               → REGRESSION: DB queries 11 (was 4). Bundle +38KB. Lighthouse 71 (was 83). All HIGH.

06-Performance → Routes to 01-PM with regression report.

01-PM         → Creates ticket. Routes to 03-SrDev with context from orbit/06.

03-SrDev      → Brain Prime: loads orbit/06 regression context + orbit/03 past performance fixes
               → Fixes: N+1 in get_posts() loop → single WP_Query with post__in
               → Fixes: tree-shaking config for scroll-animation bundle
               → Routes back to 06-Performance

06-Performance → Re-run. DB queries: 3. Bundle: +2KB. Lighthouse: 86. All pass.
               → Updates orbit/06-performance/nexterwp/budget for v2.5 baseline
               → Routes to 05-UAT

Scenario 5 — Competitor ships a feature → CTO brief → PM decision

00-CTO        → Monthly competitor pulse (via /orbit-pm-competitor-pulse)
               → Elementor Kit shipped: "AI Copilot inside block editor"
               → Assesses: High opportunity — our users want this too.

00-CTO BRIEF — Elementor Kit AI Copilot
  Signal:     Kit shipped AI block generation inside editor. WP.org reviews +320 this week.
  Assessment: Medium threat — users already asking in NexterWP support.
  Recommendation: Differentiate, not copy. Our angle: AI block config, not AI block generation.
  Owner:      01-PM runs RICE. 03-SrDev estimates effort.
  Confidence: Medium

00-CTO        → Ingests to orbit/00-cto:
                 [cto, competitor, elementorkit, ai-copilot, differentiate-with-config, 2026-05]

01-PM         → RICE: Reach 9 / Impact 7 / Confidence 5 / Effort 7 → score 45 → Q3 roadmap
               → Routes to backlog. Monitors competitor reviews monthly.

Skills → agents — who uses what

Every agent invokes specific Orbit skills. The routing is declared in routes/routes.yaml. Quick reference:

Agent Key skills they invoke
02 — Code Reviewer /orbit-wp-standards /orbit-elementor-compat /orbit-gutenberg-dev /orbit-compat-matrix
03 — Senior Dev /orbit-wp-standards /orbit-scaffold-tests /orbit-block-json-validate /orbit-i18n
04 — Dev Designer /orbit-accessibility /orbit-designer-rtl /orbit-designer-dark-mode /orbit-designer-empty-error
05 — UAT /orbit-playwright /orbit-visual-regression /orbit-user-flow /orbit-uat-gutenberg /orbit-uat-elementor /orbit-qa-regression-pack
06 — Performance /orbit-lighthouse /orbit-db-profile /orbit-bundle-analysis /orbit-editor-perf /orbit-perf-stress-test
07 — Security /orbit-wp-security /orbit-broken-access-control /orbit-sec-secrets-leak /orbit-cve-check /orbit-pay-stripe /orbit-gdpr
08 — Release /orbit-release-gate /orbit-plugin-check /orbit-release-meta /orbit-zip-hygiene /orbit-changelog-test /orbit-version-compare
09 — Docs /orbit-release-meta /orbit-i18n /orbit-pm-release-notes /orbit-abilities-api /api-documentation

Full routing: routes/routes.yaml


Always-on agents (Phase 2)

Agent files support two operating modes:

  • Mode A (now) — Operator-invoked in Claude Code. Open an agent, describe the task, it runs its SOP.
  • Mode B (Phase 2) — API runner Autonomous runner. 9 AM–6 PM IST. Autonomous scheduled dispatch. No agent file changes needed.

When Phase 2 activates, 5 always-on agents will run on schedule: 00-CTO (competitor pulse weekly), 01-PM (daily sprint routing), 06-Performance (benchmark on every commit), 07-Security (CVE feed daily), 08-Release (release gate on tag push).

→ Full architecture: docs/BLUEPRINT-ORBIT-V3.md


Install in 60 seconds

curl -fsSL https://raw.githubusercontent.com/adityaarsharma/orbit/main/install.sh | bash

That installs:

  1. Orbit cloned to ~/Claude/orbit
  2. 10 AI agents symlinked into ~/.claude/agents/ — available in every Claude Code session
  3. 116 /orbit-* skills symlinked into ~/.claude/skills/ — agents invoke these automatically
  4. WordPress/agent-skills via npx openskills install WordPress/agent-skills (WP core's official skills)
  5. Power tools: PHPCS + WPCS + VIP + PHPCompatibility, PHPStan, Playwright + Chromium/Firefox/WebKit, Lighthouse, axe-core, WP-CLI, wp-env, wp-now, source-map-explorer, PurgeCSS

After install:

# 1. Quit Claude Code fully (Cmd+Q) and reopen — agents + skills register

# 2. Seed the starter brain (one-time, requires Admin key):
bash brain/seed-brain.sh --key <orbit-admin-key>

# 3. Talk to an agent:
"UAT audit ~/plugins/my-plugin for v2.5"
"Security scan the new AJAX handler in settings.php"
"Run release gate for my-plugin v2.5"

# Or use skills directly (no brain key needed):
/orbit-do-it ~/plugins/my-plugin

What's the difference — agents vs skills?

Agents Skills
What they are SOP-driven specialists. Read brain, follow process, ingest findings. Markdown instructions — Claude runs bash/PHP/Playwright
How you invoke Natural language: "UAT audit this plugin" Slash command: /orbit-playwright
Skills vs agents Agents invoke skills automatically Skills are tools — you or an agent calls them
Brain access Yes — reads history, ingests findings No — stateless per invocation
When to use When you want the full workflow done right When you want one specific check

Use agents for releases. Use skills for quick one-off checks during development.

Update later

/orbit-update          # refreshes both agents + skills, ~20 seconds

From a clone (offline-capable)

git clone https://github.com/adityaarsharma/orbit ~/Claude/orbit
cd ~/Claude/orbit
bash install.sh

The brainless team agent

The whole vision distilled into one command:

/orbit-do-it ~/plugins/my-plugin

What happens:

  1. Auto-detects plugin type — Elementor addon, Gutenberg block plugin, WooCommerce extension, form plugin, membership/LMS, theme, or generic
  2. Picks the right pipeline — core 6 audits + type-specific add-ons + UAT + live security feeds + perf + a11y + i18n
  3. Runs in parallel with CPU throttle (auto-detects M1 / M2 / workstation)
  4. For UAT — uses /orbit-uat-agent (Stagehand-style natural-language tests; no selectors to write)
  5. Generates the master HTML report + a one-page TL;DR
  6. VerdictSHIP, WARN, or BLOCK with the top 3 things to fix

Total: ~10–15 minutes, zero questions after the path. Designed for non-technical team members + dev leads who want the audit done, not configured.

$ /orbit-do-it ~/plugins/my-new-plugin

🪐 Detected: Elementor addon (PHP 8.1+, 14 widgets)
   Pipeline: 6 core audits + Elementor (dev/controls/compat/skins/V4)
             + UAT (natural-language) + live CVE feeds + Lighthouse
   ETA: 12 min.

[12 min later]

✅ Verdict: BLOCK release — 2 Critical findings.

   Top 3 to fix:
   1. Settings page — XSS in ?search= (active probe found it)
   2. widget-3 — render() echoes attribute without esc_html
   3. widget-7 — insert time 1.4s (target < 300ms)

   Full report: ~/plugins/my-new-plugin/reports/index.html

Want even less friction? /orbit-uat-agent alone — describe flows in English ("log in → open Settings → fill API Key → save → verify saved"), the agent generates Playwright + AI-resolved selectors, runs them, self-heals on UI changes. ~$0.01–0.05 per test. Designed so a designer or PM can run UAT without writing a selector.


Runtime-evergreen, the philosophy

Software-quality tooling shouldn't freeze in the year it was written. WordPress, Elementor, Stripe, the CVE landscape — all evolve continuously. A skill that hardcodes "use apiVersion 3" is a time bomb.

Orbit's pattern, top of every SKILL.md:

## Runtime — fetch live before auditing (DO THIS FIRST)

When this skill is invoked:

1. Fetch in parallel (these are source-of-truth):
   - https://elementor.com/pro/changelog/
   - https://developers.elementor.com/docs/deprecations/
   - https://github.com/elementor/elementor/releases

2. Synthesize current state:
   - "What's the current major Elementor version as of today?"
   - "What APIs were deprecated in the last 2 minor releases?"

3. Audit against synthesized current rules — NOT against embedded text below.

4. Cite, in every finding: source URL + fetch timestamp.
   Example: `Per elementor.com/pro/changelog (fetched 2026-04-30 14:32 UTC):
            foo() deprecated in 3.22.`

That section is executable instructions for Claude, not documentation. When the skill runs, Claude reads it → fetches → uses live data.

Old pattern (snapshot) Runtime-evergreen (v2.7)
/orbit-elementor-compat "Test 3.18 / 3.20 / 3.22 / latest" hardcoded Fetches changelog → tests latest 3 minors of TODAY
/orbit-host-kinsta "Banned plugins as of April 2026" Fetches Kinsta's banned-plugins page on every run
/orbit-cve-check Pulls NVD weekly via cron Pulls NVD + Patchstack + WPScan + GitHub Advisory + MITRE per invocation
/orbit-pay-stripe "Use PaymentIntents API" (today's recommendation) Fetches Stripe API ref → uses today's recommendation

WebFetch caches for 15 minutes, so back-to-back runs in /orbit-do-it don't fire 100 fetches — unique URLs are de-duped + reused. Total overhead: ~10–30 sec on cold cache, sub-second after.

If WebFetch fails (no network), every skill has ## Embedded fallback rules for offline mode + a clear ⚠ Live source fetch failed — using fallback. Findings may be stale. notice.

Full pattern: EVERGREEN.md. Drift-checks across the suite: /orbit-skill-improver --check (action-mode meta-skill that fetches all skills' sources, diffs rules, opens PRs).


The 116 Orbit skills

Category Count Sample
Master + Brainless 4 /orbit /orbit-do-it /orbit-skill-add /orbit-skill-improver
Setup & Environment 6 /orbit-setup /orbit-update /orbit-install /orbit-docker-site /orbit-wp-playground /orbit-pre-commit
Pipeline 3 /orbit-gauntlet /orbit-release-gate /orbit-multi-plugin
Code Audits 14 /orbit-wp-{standards,security,performance,database} /orbit-{accessibility,i18n,code-quality,pm-ux-audit,compat-matrix,cve-check,abilities-api,rtc-compat,broken-access-control,scaffold-tests}
Gutenberg / Block Editor Dev 8 /orbit-gutenberg-dev /orbit-block-{render-test,edit-test,patterns,bindings,variations} /orbit-fse-test /orbit-interactivity-api
Elementor Dev 6 /orbit-elementor-{dev,controls,compat,pro,skins,dynamic-tags}
UAT Templates + Agent 6 /orbit-uat-agent (natural-language) + /orbit-uat-{elementor,gutenberg,woo,forms,membership}
QA Specialised 5 /orbit-qa-{flaky-detector,mutation,coverage,snapshot-cleanup,regression-pack}
PM Specialised 5 /orbit-pm-{rice,release-notes,feedback-mining,roadmap,competitor-pulse}
Designer Specialised 5 /orbit-designer-{tokens,empty-error,icons,rtl,dark-mode}
Browser Testing 4 /orbit-playwright /orbit-visual-regression /orbit-user-flow /orbit-conflict-matrix
Performance 7 /orbit-{lighthouse,editor-perf,db-profile,bundle-analysis} /orbit-perf-{stress-test,memory-leak,cdn}
Comparison 4 /orbit-{uat,version,competitor}-compare /orbit-changelog-test
Release 5 /orbit-{release-meta,zip-hygiene,plugin-check,block-json-validate,reports}
WP Edge Cases 7 /orbit-{multisite,uninstall-test,gdpr,cron-audit,cache-compat,rest-fuzzer,ajax-fuzzer}
Lifecycle 3 /orbit-life-{activation,upgrade,rollback}
Hosting Compat 5 /orbit-host-{wpengine,kinsta,cloudways,shared,pantheon}
Plugin Compat 5 /orbit-compat-{yoast,rankmath,wpml,polylang,acf}
Payment Integration 4 /orbit-pay-{stripe,paypal,edd,freemius}
Security Specialised 3 /orbit-sec-{xss-active,supply-chain,secrets-leak}
EU CRA + Premium 2 /orbit-vdp (EU mandate) /orbit-premium-audit (Patchstack: 76% Pro vulns exploitable)
SEO 3 /orbit-seo-{schema,sitemap,page-speed}

Full skill reference with trigger phrases + descriptions: SKILLS.md.


Composition with WordPress/agent-skills

WP core ships its own AI agent skills via WordPress/agent-skills (announcement, January 2026). The flagship skill is wp-playground — spins up WordPress in seconds via Playground CLI, gives AI agents a fast feedback loop for code iteration.

Orbit wraps; it doesn't reinvent. install.sh runs npx openskills install WordPress/agent-skills automatically. /orbit-wp-playground is a thin doc-only skill that points at WP core's runtime primitives.

Concern Owned by
Spin up WordPress for testing WP core (wp-playground)
Plugin code-quality audit Orbit (/orbit-wp-standards etc.)
Natural-language UAT Orbit (/orbit-uat-agent)
Live security feeds Orbit (/orbit-cve-check)
Multi-version matrix Orbit (/orbit-compat-matrix)
WP 7.0 Abilities API WP core runtime + Orbit audit (/orbit-abilities-api)

When WP core ships more agent skills, Orbit picks them up via the same npx openskills install chain — no Orbit code change needed.


Vision

Why this exists

Most WordPress plugin issues that reach users fall into five categories:

  1. Code that was never wrong, just untested — a widget that renders fine on the dev's machine breaks on PHP 8.2 or with WPML active or on Kinsta's edge cache
  2. Performance regressions nobody noticed — a new feature adds 40 extra DB queries per page load, or 80KB to the bundle
  3. Design debt — settings UI that confuses users because it was built dev-first, not user-first
  4. Flow blindness — nobody mapped whether a first-time user can actually complete setup without a tutorial
  5. No comparison baseline — "our Mega Menu is better than ElementKit" stated without any data

UAT (User Acceptance Testing) is the practice of validating a product from every perspective before it ships — not just "does the code run" but "will a real user get stuck, is the UI regressed, does the PM have evidence it's better than competitors." Orbit automates that entire layer for WordPress plugins.

What top teams do that most don't

  • Automattic / WordPress VIP run every commit through PHP linting + VIP coding standards before merge
  • 10up uses AI-powered visual regression — catches when something looks different without being technically broken
  • WordPress.org plugin team added 15+ automated security checks in 2025 alone
  • Leading Elementor addon teams run Playwright E2E suites across 3 WP versions before release

Orbit brings that same discipline to any plugin team, with a single command.

The three rules

  1. Local-first, not CI-first. Real MySQL, real PHP, real browsers — already on your Mac. CI is optional plumbing.
  2. Skills are senior reviewers, scripts are junior QA. Claude Code skills read the code the way an experienced senior developer would. Scripts handle deterministic checks.
  3. Skills must be runtime-evergreen. No quarterly maintenance. Every skill fetches its canonical source on every run.

What's coming next

  • WP 7.0 readiness (ships May 20, 2026) — already covered by /orbit-abilities-api + /orbit-rtc-compat + the runtime-fetch pattern
  • EU Cyber Resilience Act compliance/orbit-vdp is mandatory; /orbit-premium-audit covers the 76% premium-exploitability gap
  • Elementor V4 Atomic (default for new sites April 2026) — /orbit-elementor-compat auto-handles via runtime-fetch
  • Cloud-hosted runs (orbit.run, future) — gauntlet on a PR via GitHub Action, no local Docker
  • Community contributions/orbit-skill-add is a meta-skill that scaffolds new skills in the Orbit pattern. Anyone can add a skill via PR; the community catalogue grows.

Severity model

Every Orbit skill applies this triage:

Level Action before release
Critical Block release. Fix immediately.
High Block release. Fix in this PR.
Medium Fix if under 30 min. Otherwise log + defer.
Low / Info Log in tech debt. Defer.

/orbit-do-it reads these consistently and produces a single SHIP / WARN / BLOCK verdict at the top of every report.


Reports

Every audit run drops everything into reports/:

reports/
├── qa-report-<timestamp>.md           ← markdown summary
├── tldr-<timestamp>.md                ← one-page verdict
├── index.html                         ← master HTML (PM-friendly)
├── playwright-html/index.html         ← visual test report
├── skill-audits/index.html            ← tabbed AI audit
├── uat-report-<timestamp>.html        ← UAT comparison + videos
├── pm-ux/pm-ux-report-*.html          ← PM-friendly UX report
└── lighthouse/lh-<timestamp>.json     ← Core Web Vitals

Open the master index:

open ~/plugins/my-plugin/reports/index.html

Designed to be shared with PMs / managers / customers without terminal access.


Standards this follows


Contributing

Open to:

  • New skills — fork, run /orbit-skill-add, follow the runtime-evergreen pattern, open a PR
  • Skill improvements — every skill has Sources & Evergreen References. If a source moved or a rule needs updating, /orbit-skill-improver --pr opens a draft for review
  • Edge-case reports — file a GitHub issue with [skill] or [bug] tag and a minimal repro

Keep contributions research-first. Every check should link to the standard or incident that motivated it.


Built by

Aditya Sharma · POSIMYTH Innovation github.com/adityaarsharma/orbit

The discipline: Software-quality tooling shouldn't freeze in the year it was written. It should know what today looks like by re-reading the canonical sources every time it runs. That's runtime-evergreen. That's Orbit.

About

🪐 10-agent WordPress Plugin QA team for Claude Code. 116 runtime-evergreen skills · CTO→PM→Dev→UAT→Security→Release · MCP discovery · Docker wp-env · WP.org zero-rejection · Brain-connected memory.

Topics

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors