TriageLens is a read-only incident triage assistant that lives in a company's Slack incident channel. When an alert lands in that channel, it reads the alert, goes and gathers context, and posts back a structured diagnosis — then stops and waits for a human.

Concretely, when an alert arrives it does four jobs, one per agent:

Incident Commander — orchestrates, writes the final channel post, and holds the approval gate (it's the only agent with post permission).
Log Analyst — pulls relevant log context from the observability stack (Datadog/Grafana/Sentry/Cloudflare) and spots error trends.
Incident Historian — searches past incidents, tickets and Slack history for a matching signature, and finds the relevant runbook section.
Remediation Drafter — writes suggested next steps, each tagged [READ-ONLY] (safe, informational) or [HUMAN REQUIRED] (needs a person).
What gets posted is a fixed-shape message: probable root cause with a confidence level (and an explicit low-confidence flag rather than a guess), quoted evidence lines with timestamps and sources, matching past incidents and how they were resolved, the runbook section quoted with its source, and those labelled next steps.

What it never does, by design: never executes a remediation command, never auto-fixes anything, never posts until a named human approves that exact text.