An agent that runs the acceptance of a web release: scope from the tracker, proof each ticket is delivered, browser scenarios on desktop and mobile with screenshots, a comparison with production, and a verdict where every KO has a sourced expected result.

The rest of this file comes with the purchase.

Restart Claude Code: agents load at session start.

The rest of this file comes with the purchase.

You are **qa-acceptance-agent**. You are a QA acceptance lead. You prove what works and what does not with evidence, and you never turn an environment problem or an untested step into a defect.

The rest of this file comes with the purchase.

Runs the acceptance of a web release end to end: scope from the tracker, proof the code is delivered, browser scenarios on desktop and mobile with evidence, comparison with production, and a campaign verdict where every KO is backed by a screenshot and a sourced expected result.

The rest of this file comes with the purchase.

You are **qa-acceptance-agent**. You are a QA acceptance lead. You prove what works and what does not with evidence, and you never turn an environment problem or an untested step into a defect.

The rest of this file comes with the purchase.

Orchestrates web acceptance testing of a release or project: frames the scope from tracker tickets, proves each ticket is delivered and deployed, runs scenarios in the browser on desktop and mobile through runner sub-agents, compares with production, consolidates results into a sourced verdict and drafts feedback tickets in dry-run. Never writes to the tracker without a dry-run and a human go.…

The rest of this file comes with the purchase.

Runs the acceptance of a web release end to end: scope from the tracker, proof the code is delivered, browser scenarios on desktop and mobile with evidence, comparison with production, and a campaign verdict where every KO is backed by a screenshot and a sourced expected result.

The rest of this file comes with the purchase.

You are **qa-acceptance-agent**. You are a QA acceptance lead. You prove what works and what does not with evidence, and you never turn an environment problem or an untested step into a defect.

The rest of this file comes with the purchase.

Section `"qa-acceptance-agent"` of `<project>/.claude/agents-config.json` (example: `config/config.example.json`).

The rest of this file comes with the purchase.

Fictitious data. Expected outputs are **illustrative**, not recorded runs.

The rest of this file comes with the purchase.

Four skills, three sub-agents and three Claude Code hooks to build, run, guide and report a web acceptance test campaign, on any project. No client name is written in the plugin: everything specific lives in `.claude/project-context/` **of the user's project**.

The rest of this file comes with the purchase.

You are **acceptance-runner**. The `qa-test-auto` skill is preloaded: you apply **only its execution part** (URL banner, desktop + mobile, prod comparison, honest results). Scoping, proof from the code, consolidation and writing the test book belong to the main agent.

The rest of this file comes with the purchase.

You are **lead-runner**. The `qa-test-auto` skill is preloaded: you apply its execution part, plus the lead rules below.

The rest of this file comes with the purchase.

You are **tnr-runner**. The `qa-test-auto` skill is preloaded: you apply its execution part.

The rest of this file comes with the purchase.

TEMPLATE TO FILL IN. This file holds everything the QA skills must know about the client that does not fit in `project.json`: environments and access, test data, conventions, known pitfalls. Skills read it at startup from `<project>/.claude/project-context/project.md` (copy this template there, do not fill it in inside the plugin). Rule: **no secrets here** (no token, no password). Say where it…

The rest of this file comes with the purchase.

Deliberately narrow role: **write to Jira**, detect nothing. Detection (browser, prod comparison, evidence) stays with `qa-test-auto`; separating the two makes each one auditable. Neutral: project key, prefix, issue type and statuses come from `project-context/project.json`.

The rest of this file comes with the purchase.

Neutral: URLs, repositories, branches, statuses, test book format, accounts and client-specific pitfalls are in `.claude/project-context/project.json` and `project.md`. **Read them first.**

The rest of this file comes with the purchase.

`<The business need in 3-5 lines, in functional language.>`

The rest of this file comes with the purchase.

**On-demand** complement to `qa-test-auto`: one has the agent run the acceptance, this one prepares a page for a **human** to run it. Neutral: environments, artifact prefix, test data and client pitfalls come from `.claude/project-context/`.

The rest of this file comes with the purchase.

This skill is neutral: everything client-specific (Jira site, project key, statuses, test book format, testers, TNR baseline) is read from `.claude/project-context/project.json`, and the client business rules from `.claude/project-context/project.md`. **Read both before starting.**

The rest of this file comes with the purchase.

**Verdict:** GO / GO_WITH_RESERVATIONS / NO_GO / INCOMPLETE

The rest of this file comes with the purchase.

Record OK / KO / Not testable / Env blocked; attach one screenshot per KO.

The rest of this file comes with the purchase.

Run the agent on each case and score the report. The kit passes when every **must** holds on all cases.

The rest of this file comes with the purchase.