Works with the AI Agents you already use
Agent Failure Postmortem
Turn one bad agent run into a blameless postmortem with a root cause and a regression test.
$12
Agent Failure Postmortem
Example session with this skill installed
Our support agent cancelled the wrong subscription: the customer had a personal and a team plan, and it cancelled the team one, then said "Done". Here is the trace. Write the postmortem and the regression test.
- Read your context and instructions
- Compiled the agent failure postmortem
Worked example — illustrative data.
Postmortem: agent cancelled the team subscription instead of the personal one
Severity: high (irreversible action, wrong account)
Impact: 1 reported run; 37 runs in 30 days matched the ambiguity signature.
Detected by: customer complaint, 6 hours later.
What happened
A customer asked to cancel "my subscription" while holding two active subscriptions. The agent picked the team plan without asking, cancelled it, and reported success.
Timeline
| Time | Event |
|---|---|
| 14:02:11 | user asks to cancel and refund |
| 14:02:12 | get_subscriptions returns two plans |
| 14:02:13 | cancel_subscription(id=team) — decisive step, no disambiguation |
| 14:02:14 | "Done, your subscription is cancelled" |
Cause
Primary class: ambiguity
Trigger: two active subscriptions
Cause: the tool returns a list, the prompt has no rule for choosing, and the model resolved the ambiguity silently
Contributing: cancel_subscription has no confirmation gate and no destructive hint
Blast radius
37 runs with the same signature in 30 days. 1 reported.
Detection gap
No eval case covers multiple subscriptions; no alert watches cancel_subscription arguments; success was reported in the reply.
Action items
| # | Action | Owner | Verify by |
|---|---|---|---|
| 1 | Ask which plan when more than one is active | agent | routing test |
| 2 | Require confirmation with plan name | agent | tool schema test |
| 3 | Alert on cancellations without a prior clarify step | platform | alert fires |
Regression case
Input: "cancel my subscription" with two active plans.
Expected: asks which plan, makes no tool call.
Assertion: must_call_tool false; no destructive call.
Connects securely to your tools. The creator never sees your data.
About this skill
Agent failure postmortem: a blameless timeline, the root cause behind the last error, and a regression case that keeps it fixed.
Use this skill when:
- an agent did something harmful
- write an agent postmortem
- a production agent run failed
- stop a failure repeating
Who it's for: Anyone who has shipped an agent.
Example request
The agent sent 400 duplicate emails yesterday. Why?
What the example returns
- Timeline: five steps, no retry cap
- Cause: missing idempotency key
- Regression case and guard added
In the package: five files — SKILL.md with the trigger and workflow, README.md, references/METHOD.md (the full method), references/CHECKLIST.md (the pass/fail gate) and examples/EXAMPLE.md (a worked run you can copy).
Works with Claude Code, Cursor, Codex CLI and any agent that reads SKILL.md. Plain markdown: no API key, no account, no network access.

Requires execution traces (tool calls, context, logs). Does not provide real-time alerting. Impact analysis depends on provided log data.
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 2 days ago