One loop, aimed at your hardest work.

Keep production green

Incidents fix themselves before you read the alert.

Errors pile up faster than anyone can triage them, and the fix is usually small but nobody has the context at 2am.

1. 1

   Monitor polls Sentry and PostHog and ranks issues by blast radius.

2. 2

   A cycle reproduces the bug, writes the fix, and adds the missing test.

3. 3

   VERIFY boots the app and polls its health URL — “compiles” isn’t “works”.

4. 4

   You approve the merge gate; Aion deploys and watches for regressions it caused.

Burn down the backlog

Ship features from a sentence of intent.

The backlog is full of clear, scoped work that just needs hands. Writing prompts step-by-step is slower than doing it yourself.

1. 1

   Give the outcome — “users can export reports as CSV” — not the steps.

2. 2

   PLAN names a branch, CODE writes the change and the test that proves it.

3. 3

   SHIP opens a PR whose body says what, why, and how it was verified.

4. 4

   Run one at a time, or let `aion loop` work the whole queue.

Incident → PR, autonomously

A production error becomes a merged pull request.

The path from “users hit an error” to “fix is live” is full of handoffs — alert, triage, assign, context-load, code, review, deploy.

1. 1

   Sentry groups the error into one issue with a stable fingerprint.

2. 2

   A mission is enqueued with full context; duplicates never re-open.

3. 3

   The cycle fixes, tests, ships, and (after your gate) deploys.

4. 4

   Failures loop back to CODE — a fix is never shipped unverified.

Migrations & sweeps at scale

Mechanical change across the whole repo, verified each time.

Framework bumps, API renames, and codemod-style sweeps are tedious, risky, and easy to half-finish.

1. 1

   Scope the change once; Aion runs a cycle per target, isolated.

2. 2

   Every change is tested and the app is booted before it ships.

3. 3

   A 24-step budget ends stuck cycles; a 3-cycle failure is dismissed with a reason.

4. 4

   Everything is files on disk — inspectable, git-trackable, restart-safe.

Security hardening on autopilot

DDoS defense, vuln scanning, secrets audits — preinstalled.

Security work is important and never urgent, so it slips. Findings need someone to actually ship the remediation.

1. 1

   Ships with security skills: DDoS defense, vuln scanning, secrets audits, edge hardening.

2. 2

   Phase output is scrubbed for secret shapes (tokens, AWS keys, JWTs) before storage.

3. 3

   Findings become missions; remediations run as gated cycles.

4. 4

   Disabled skills stay out of the agent’s context — you choose the surface.

Stay shipping while you sleep

A queue that works itself, overnight and on.

Throughput is capped by waking hours. Most of the cycle doesn’t need a human — only the gates do.

1. 1

   `aion loop` polls signals, takes the highest severity, runs a cycle, repeats.

2. 2

   Gates pause the loop for one-click approval; everything else is autonomous.

3. 3

   Each rung of the autonomy ladder keeps the same audit trail.

4. 4

   Scheduled automations triage on a cron with the bundled runtime.