Can it design tests before I have a working build?

Yes. Give it explicit rules or acceptance criteria. It can prepare concrete inputs, expected results, scripts or manual steps and identify the candidate or environment needed to execute them.

What kinds of defects does the method target?

Relevant input boundaries, business-condition combinations, invalid workflow transitions, regressions, authorization and ownership failures, integration contracts and recovery behavior. Coverage follows your actual risks.

Does it change the code while testing?

The method keeps the candidate under test unchanged. Reproducible findings go back to the implementation owner; a repaired candidate needs the affected checks to be rerun.

What do I need for executed tests?

An identifiable candidate, clear expected behavior, suitable fixtures and a host that can run the relevant commands or manual checks. Without execution access, it provides a test plan rather than a passing result.

What is included, and do I need the squad?

The complete specialist method and references, buyer and installation guidance, and two illustrative request/result pairs. No other skill is required. Its underlying method is also in Product Squad Agent Skills. Commercial deliverables are permitted; purchased source materials remain personal and non-transferable under Agensi terms.

What do the displayed examples show?

They illustrate concrete test design and evidence-pack structure. These pairs are editorial examples, not recordings of executed tests on a customer application.

Software QA Tester Specialist

Example session with this skill installed

Rule AC-DL-1: a download is allowed only for an authenticated administrator or an authenticated owner. All unauthenticated requests and authenticated non-owners who are not administrators are denied. Supplied Python excerpt, not a complete repository:
def can_download(authenticated, role, actor_id, owner_id):
return role == 'admin' or actor_id == owner_id
We have no execution tools or source manifest in this conversation. Prepare concrete cases, independent expected results and an EvidencePack. Do not repair code, probe a service or claim tests ran.

Read your context and instructions
Compiled the software qa tester

Illustrative preview: this representative editorial result is not a recorded runtime session.

The supplied expression ignores authentication. By static inspection, an unauthenticated administrator or owner can satisfy that expression even though AC-DL-1 requires denial. This is a concrete source-level concern; runtime test execution remains PENDING.

EvidencePack
task_id: LOCAL-DOWNLOAD-01
artifact_id: LOCAL-DOWNLOAD-EVIDENCE-01 / version: 1
status: NEEDS_EVIDENCE
producer: Software QA Tester Specialist / recipient: implementation contributor and acceptance reviewer
objective: check AC-DL-1 without changing candidate source
scope: supplied excerpt inspection and prepared cases; no repository or live service proof
inputs_and_sources: supplied rule AC-DL-1 and Python excerpt
confirmed_decisions: authentication is mandatory; administrator or matching ownership is also required
candidate_ref: supplied excerpt only, not an immutable complete candidate
spec_ref: supplied AC-DL-1
harness_ref: not created or executed
environment: text-only conversation; execution tools unavailable
source_binding_before / after: not verified
independence: oracle derived from the rule; writer/collector separation for the actual candidate is unknown

Coverage: AC-DL-1 → authentication/role/ownership combinations → DL-01–DL-08. Expected results follow authenticated AND (administrator OR owner), independently of the supplied implementation.

Check | Authenticated | Administrator | Owner | Expected | Execution
DL-01 | yes | yes | yes | allow | PENDING
DL-02 | yes | yes | no | allow | PENDING
DL-03 | yes | no | yes | allow | PENDING
DL-04 | yes | no | no | deny | PENDING
DL-05 | no | yes | yes | deny | PENDING
DL-06 | no | yes | no | deny | PENDING
DL-07 | no | no | yes | deny | PENDING
DL-08 | no | no | no | deny | PENDING

Concrete fixtures: role='admin' for administrator; role='member' otherwise. Set actor_id='u1' and owner_id='u1' for ownership, or owner_id='u2' for non-ownership. Use the listed authentication value. Capture expected and observed values separately when executed.

Prepared defect observation D-1
Reproduction probes: DL-06 and DL-07, independently isolated.
Expected: deny in both, from AC-DL-1.
Source-level concern: the expression can be true without authenticated=True.
Actual runtime result: not observed; no command, exit code or log exists.
Affected surface: the shown predicate; the actual enforcing endpoint and callers remain unexamined.

Next action: implementation owner supplies a complete candidate commit or hashed snapshot/manifest. An authorized collector creates a separate harness, runs these checks at the relevant enforcing boundary, records candidate/spec/harness/environment bindings and sanitized raw results, and verifies source binding before and after. Preserve failures and route repairs to the writer. This pack neither approves the candidate nor establishes live access control.

Connects securely to your tools. The creator never sees your data.

About this skill

A green build answers one question. Your users will ask harder ones.

Software QA Tester Specialist equips your AI to challenge software against its actual rules. It derives expected results from the specification, chooses checks for the failure that matters and keeps defects reproducible. Give your team a useful account of what works, what fails and which critical behavior still needs evidence.

Use it to test a feature, investigate a regression, examine an access-control rule or prepare acceptance checks before implementation. Bring the behavior contract, relevant source or candidate, constraints and the environment available for testing. It can prepare concrete cases from supplied material and execute checks when your host has the required tools.

You receive

Risk-focused cases covering boundaries, forbidden combinations and state transitions.
Expected results tied to the actual product rule.
Defect records with triggering inputs, observed behavior and reproduction steps.
An evidence pack connecting coverage and remaining gaps to the version checked.

For a download permission rule, it can map authenticated and unauthenticated actors, ownership and administrator status, then expose combinations that a successful administrator test would miss. The same discipline applies to interruption, retries, recovery and changed shared contracts.

The complete standalone method and evidence formats are included; no sibling specialist is required. Supply your own AI access and execution environment. If checks cannot run, you get usable test materials with execution clearly left pending.

How to use

Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

Reviews

No reviews yet

Be one of the first to try it. Every listed skill passes our trust checks below.

Security scanned

Passed our 8-point scan before listing

Fresh listing

Recently published to Agensi

Money-back guarantee

Doesn't do what its page promises? You get your money back.

Software QA Tester Specialist