- Home
- Skills
- Workflow Automation
- Automation Replay Gate — Workflow Failure QA
Works with the AI Agents you already use
Automation Replay Gate — Workflow Failure QA
Find duplicate and recovery failures before enabling an automation. See what each replay created and what needs fixing.
Free
Automation Replay Gate — Workflow Failure QA
Example session with this skill installed
Actual synthetic demonstration. The supplied event-to-task-to-notification workflow and local JSON stores are disposable; no task service or notification service is connected. They are author-side demo fixtures, not bundled product components.
Before enabling workflow v0.1, use Automation Replay Gate to run evt-demo-001 normally, deliver it twice, lose the acknowledgement after the task commit, and fail notification delivery before resuming from saved state. Expect exactly one task and one notification per event. Show the observed effects. You may repair this local fixture and replay it, but do not activate anything.
- Read your context and instructions
- Compiled the automation replay gate
Recovery bug found and repaired in the local fixture
Actual synthetic runs, 21 October 2026 at 05:09:28 UTC. Both external effects were replaced with local JSON stores. No n8n engine, MCP service or live provider ran.
Expected final state for each case: one task and one notification.
| Replay | v0.1 tasks / notifications | v0.2 tasks / notifications |
|---|---|---|
| Normal event | 1 / 1 | 1 / 1 |
| Same event delivered twice | 1 / 1 | 1 / 1 |
| Timeout after task commit | 1 / 1 | 1 / 1 |
| Notification failure, then resume | 1 / 0 — failed | 1 / 1 — passed |
Cause: v0.1 returned DONE as soon as it found the existing task. The resume path never checked whether the notification had been sent.
Repair: remove that early return. New and existing tasks both continue to notification readback. The saved task is reused; the missing notification is sent once. The four replays then matched their declared effect counts.
Verdict: REVISE for v0.1; PASS for v0.2's isolated replay. The retained evidence includes both workflow versions, the patch, ordered effect journals and saved final states.
Next step: review the actual provider's duplicate and acknowledgement contracts, then run an authorized live smoke test. The workflow remains inactive.
Connects securely to your tools. The creator never sees your data.
About this skill
A workflow creates the task, then its notification fails. On the next run it sees that the task exists and stops. The task is there; the notification is still missing.
Automation Replay Gate helps you find that kind of unfinished work before enabling a workflow. Give your agent the actual workflow and a disposable runner. It identifies the expected effects, checks that every outbound path is isolated, then exercises normal delivery, duplicates, a timeout and a downstream failure followed by recovery.
The result is a replay table showing expected and observed effects, the failed trace and the next repair. You can see whether a retry repeats work, skips a remaining step or reaches the intended final state.
Use it before the schedule goes on
It fits a scheduled batch, webhook, n8n flow or agent workflow whose effects and recovery rules can be inspected and replayed. You supply the runnable harness or authorized sandbox. The free skill provides the investigation and replay procedure; it includes no workflow engine, n8n integration or universal runner.
Missing isolation or an unclear contract stops execution with the exact missing requirement. A passing isolated replay covers those fixtures. Check the live provider and authorize activation separately.
Start with the workflow file and one example event.
The buyer supplies a runnable workflow, material effect contracts and an isolated harness or authorized sandbox.
Every outbound path must be isolated; a control named dry run is insufficient by itself.
Unclear contracts, missing observation or uncertain isolation produce HOLD.
PASS covers executed fixtures, not universal exactly-once delivery, live provider behavior or activation authorization.
The demo workflow and test harness are author-side evidence and are not bundled into the existing skill.
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
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 24 days ago
- Free with an Agensi account
Permissions
Read workflow files and execute an isolated local or authorized sandbox replay. Workflow edits and sandbox writes follow the task scope. No live activation, external send, upload, or production account write is required for the core replay.