- Home
- Skills
- Workflow Automation
- Lazy Test: The Smallest Real Proof
More screenshots
Works with the AI Agents you already use
Lazy Test: The Smallest Real Proof
Verify changes with the smallest real proof: rank checks by risk and fan-out, run the minimal set, interpret honestly, then stop.
$5
Lazy Test: The Smallest Real Proof
Example session with this skill installed
I changed the token-refresh logic in the auth module. Prove it works — and tell me what you did NOT run and why.
- Read your context and instructions
- Compiled the lazy test
Ranked before running: the changed contract is token refresh, exercised the way callers actually use it — that is non-negotiable. Direct dependents are the session middleware and the login flow. The affected suite is the auth module. The full suite was not run: the change is module-local, no shared code touched, so the full suite would be theater. Proof matched to the change type: unit tests covering the new behavior plus expiry edge cases, run against the real handler with error paths included — expired token, revoked token, concurrent refresh. RESULT: 47/47 pass across the auth module and its dependents. Interpreted honestly: pass means pass on the code as it stands — no reassurance re-runs. One test failed on the first run and passed on the second with no code change; it was logged as a flake and quarantined, not celebrated as a fix. Report, terse: RESULT 47/47 pass. COVERAGE — changed contract (token refresh) exercised including expiry edge cases; full suite not run, change is module-local. RISK — none outstanding; migration path not applicable. Named explicitly, not hidden: NOT VERIFIED — mobile Safari rendering of the session-expired banner, no device available. That sentence is more valuable than a paragraph of confidence.
Connects securely to your tools. The creator never sees your data.
About this skill
Testing is not a ritual — it is a claim about risk. Every test you run should reduce uncertainty about something that could actually break. Run the smallest set that covers the changed contract, interpret the results, and stop when the risk is retired. Expand only when a failure or an untested risk demands it. The skill ranks checks before running them — changed contract first, then direct dependents, then the affected suite, then the full suite only for wide fan-out or shipping — and defines what counts as proof for each change type: unit tests for logic, real-handler requests for endpoints, rendered output for UI, a clean install for packaging, up-and-down migration on real-shaped data. The inequalities are the spine: lint is not runtime, build is not behavior, skipped is not passed. A second run that passes without a code change is a flake, not a fix.
It does not expand 'just to be thorough' — thoroughness without a risk behind it is how a 5-minute change eats an hour of compute.
A pre-existing failure on clean trunk is reported, not fixed inline — unless it blocks you, and then the skill says so.
'Works on my machine' means the environment is not controlled: pin it down or record versions.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by 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.
Trust & safety
Security scan
Security scan passed 3 days ago
Requires