acceptance-testing

name	acceptance-testing
description	Use when requesting-code-review is complete and all Critical/Important issues are resolved, before finishing-a-development-branch - dispatches an acceptance-tester subagent to verify every acceptance criterion

Acceptance Testing

Announce at start: "I'm using the acceptance-testing skill to verify the branch against acceptance criteria."

Dispatch superpowers:acceptance-tester subagent to verify the completed branch satisfies every acceptance criterion before integration. The tester receives the AC document path and the current git HEAD SHA.

Core principle: Implementation is not done until every acceptance criterion has recorded evidence of a pass. A failing criterion is a blocker, not a review comment.

---

The Iron Law

NO BRANCH FINISHING UNTIL ACCEPTANCE TESTING REPORTS ALL CRITERIA PASSED
(OR EXPLICITLY DEFERRED WITH USER APPROVAL).

---

When to Use

Mandatory:

• After requesting-code-review completes and all Critical and Important issues are resolved
• Before invoking finishing-a-development-branch

If no AC document exists: Stop. Invoke superpowers:writing-acceptance-criteria to produce the document from the spec, get user approval, then return here.

---

How to Run

1. Identify inputs:

AC_DOC="docs/superpowers/acceptance/YYYY-MM-DD-<feature>.md"
HEAD_SHA=$(git rev-parse HEAD)
REPO_ROOT=$(git rev-parse --show-toplevel)

2. Dispatch acceptance-tester subagent:

Use Task tool with superpowers:acceptance-tester type, fill template at acceptance-tester.md

Placeholders:

• {AC_DOC_PATH} — path to the acceptance criteria document
• {HEAD_SHA} — current git HEAD SHA
• {REPO_ROOT} — absolute path to the repository root

3. Act on the report:

Report outcome	Action
All criteria PASS	Proceed to superpowers:finishing-a-development-branch
One or more criteria FAIL	Return to superpowers:executing-plans: create fix tasks for each FAIL criterion
Criteria Blocked due to failed dependency	Will unblock automatically once their dependency criteria pass; treat as FAIL for fix prioritization
Criteria Blocked due to missing infrastructure	Stop and report to user; get explicit deferral approval before proceeding

---

On Failure: Fix Loop

When the acceptance-tester reports failures:

• Do NOT return to brainstorming
• Do NOT modify the AC document to remove or weaken the failing criterion
• Announce each failing criterion ID and description
• Return to executing-plans and create targeted fix tasks
• Re-run requesting-code-review on the fixes
• Re-run acceptance-testing after fixes are complete

The fix loop is: executing-plans → requesting-code-review → acceptance-testing. Repeat until all criteria pass.

---

Red Flags

• Invoking finishing-a-development-branch while any criterion shows FAIL
• Removing or relaxing a criterion because it is hard to pass
• Accepting "should be fine" from the subagent report as a PASS
• Skipping acceptance testing because "tests already pass" — unit tests and acceptance criteria test different things

---

Integration

Called after:

• superpowers:requesting-code-review — must complete before acceptance testing begins

Calls:

• superpowers:acceptance-tester (subagent, via Task tool)
• superpowers:executing-plans (on failure, to create fix tasks)
• superpowers:finishing-a-development-branch (on full pass)

If AC document is missing:

• superpowers:writing-acceptance-criteria — produce the document first

Required external skill:

• superpowers:playwright-cli — needed for UI interaction criteria; install from https://github.com/microsoft/playwright-cli (see the skills/ directory) if unavailable

See subagent template at: acceptance-testing/acceptance-tester.md