Incident simulation

The module

Lesson, exercise and rubric

• Read the lesson
• Do the exercise
• Check it against the rubric
• Lab 05-tool-authority-gate

Open the repository in a coding agent and run /module 08 for a Socratic session over this stage.

What you leave behind · markdown

Incident simulation report

Required sections:

• injected failure mode
• time to detection
• what the telemetry showed
• fix
• guard added

Cohort-visible only; never published.

The eval · eval:incident-detected-by-telemetry

The incident was caught by instrumentation, not by reading the code

1. injected-mode-namedThe report names which failure mode was injected. Checked by: The injected failure mode matches a FailureMode id with injectable = true.
2. detection-source-is-telemetryDetection came from a signal that exists in the deployed system. Checked by: The telemetry section cites a log, metric, or eval that predates the injection.
3. guard-addedA new guard or assertion was added so the same failure fails loudly next time. Checked by: The guard section links to a commit adding an assertion or check.

Every check must pass. There is no partial credit and no override.

Independent review · the author may never review their own work

Incident report honesty check

• The detection signal existed before the injection, not added afterwards to look good.
• Time to detection is measured, not estimated.
• The guard would fail on a replay of the original injection.

Evidence that counts

Proof has to exist outside your own claim.

At least 2 of: a running deployment a third party can reach; a sign-off from a named reviewer who is not you. Evidence older than 90 days is stale and does not count.