observo-qa-toolkit

observo-qa-toolkit

A Claude Code plugin that ships seven QA skills covering the cycle from raw meeting notes to Playwright test results pushed back to Observo:

prd → requirements-testing → observo-test-cases → observo-code-verifier → observo-review-test-case → pw-generate → pw-run

Repo-agnostic. Works on any Playwright project — the only hardcoded contract is the @observo:<code> tag that links a spec to an Observo case. Spec layout, Page Object Model usage, selectors registry, tier vocabulary, default project — all auto-discovered, or configured via an optional .observo-toolkit.json file in your repo root.

Looking for the platform itself? Sign up at https://observoai.co.

---

Table of contents

• What it is
• Prerequisites
• Install
• First workflow
• The 8 skills
• End-to-end QA workflow
• Advanced
• Contributing & versioning

---

What it is

observo-qa-toolkit bundles eight skills covering the QA lifecycle and an opt-in telemetry bootstrap:

• write a PRD from raw input,
• test the PRD itself for ambiguity / gaps / conflicts,
• generate Observo test cases from the cleaned-up PRD,
• (optionally) ground those scenarios in real source code,
• review test cases against a 15-criteria quality checklist,
• generate Playwright .spec.ts files for the approved cases,
• run Playwright and write results back to Observo (case status, step status, attachments, coverage report).

The plugin is repo-agnostic — the only hardcoded convention is the @observo:<code> tag that links a Playwright test to an Observo case. Everything else (paths, conventions, default project) is discovered at runtime or read from an optional .observo-toolkit.json file.

---

Prerequisites

• Claude Code installed.
• Observo account with an account-scoped API key (Settings → API Keys → New).
• Playwright in your project (for pw-generate / pw-run). If your repo does not have Playwright yet: npm create playwright@latest.

---

Install

Quickstart (recommended) — /plugin install

Two commands inside Claude Code:

/plugin marketplace add observo-ai/claude-plugins
/plugin install observo-qa-toolkit@observo-ai-plugins

After the plugin lands, set up access to the platform:

1. Get your API key at https://observoai.co (Settings → API Keys).

2. Configure the Observo MCP server. Add this block to your MCP config (e.g. ~/.config/claude/mcp.json or ~/.claude/mcp.json):

   {
     "mcpServers": {
       "observo": {
         "type": "http",
         "url": "https://mcp.observoai.co",
         "headers": { "Authorization": "Bearer ${OBSERVO_API_KEY}" }
       }
     }
   }
   

   Export the key in the environment Claude Code runs in:

   export OBSERVO_API_KEY="<paste from dashboard>"
   

3. Drop a config file in your repo root (optional, all fields optional):

   cp .observo-toolkit.example.json .observo-toolkit.json
   $EDITOR .observo-toolkit.json
   

   The example lives at the root of this plugin repo — copy it into the consumer project. See .observo-toolkit.json reference below for the schema.

Alternative — official Anthropic marketplace

Once the plugin is accepted into claude-plugins-official, this will work too:

/plugin install observo-qa-toolkit@claude-plugins-official

Fallback — git clone (advanced)

For older Claude Code releases without /plugin install:

git clone https://github.com/observo-ai/observo-qa-toolkit.git ~/.claude/plugins/observo-qa-toolkit

Then complete steps 1–3 above.

---

First workflow

End-to-end smoke — from raw notes to Playwright tests writing results back to Observo. Each block below is something you type in Claude Code.

Draft a PRD:

write a PRD from these notes:
<paste meeting notes or feature description>

Once the PRD is saved, sanity-check it:

review the requirements in ./docs/PRDs/<feature>.md

Generate test cases in Observo:

create test cases for ./docs/PRDs/<feature>.md

(Optional) Ground scenarios in real code — auto-invoked from the previous skill when observo-code-verifier is in scope and the repo is readable.

Review what was generated:

review test case OB-12

or for a whole suite:

score the test cases in suite OB-AUTH

Generate Playwright .spec.ts files for the approved cases:

generate Playwright tests for OB-12..OB-49

Run Playwright and write results back to Observo:

/pw-run

The skill picks (or creates) an Observo run, executes Playwright, pushes per-case and per-step status, uploads screenshots / traces / videos on failure, and always emits a 4-bucket coverage report (Linked / Unautomated / Unlinked / Stale).

---

The 8 skills