Sun Runs

Level 3 · Course 20

Sometimes you do not need another planning session. You need your agent to put its head down and run. But an unconstrained sprint turns into chaos: skipped checks, forgotten tools, old blockers rediscovered, no journal, no extraction. The agent looks busy and the system does not improve.

A sun run is focused execution with discipline built in. Your agent runs hard, but it must check the relevant pipeline before declaring blockers. It must search the vault for prior solutions. It must verify tools instead of trusting stale notes. It must QA the last output. And when the run ends, it must write down what happened and extract what should become permanent infrastructure.

This course installs the sun run playbook. You define the mission, the minimum checks, the blocker protocol, the journal, and the stop condition. Then your agent runs until the chunk is done or context says wind down. The point is not speed alone. The point is compounding speed: every sprint makes the next sprint cleaner.

A Worked Sun Run

A business owner has been avoiding a maintenance backlog for three weeks. The work is real and connected: audit the filing system, update twelve trigger word entries, rebuild the boot chain redirects, and journal everything that changed. None of it is hard. All of it needs to happen in sequence, and starting it has felt like opening a door to a mess with no clear exit.

That is exactly what a sun run is for. The agent does not guess at scope. It starts with a pre-flight check: Is the pipeline spec current? Are the filing tools reachable? Has the vault been searched for prior attempts at this same cleanup? What are the stop conditions? The agent answers each question before touching a single file. Then it runs.

Phase 1: filing audit. Agent scans the system, flags gaps, produces a structured report. Gate: operator reviews the gap list before any files are moved. Phase 2: trigger word updates. Agent opens the trigger word file, makes the twelve edits, presents a diff. Gate: operator confirms. Phase 3: boot chain redirects. Agent checks each redirect, rebuilds what is broken, logs each change. Gate: agent verifies each redirect resolves cleanly before moving on. Phase 4: journal. Agent writes a compact session log covering what was done, what changed, and one extraction — something that should become permanent infrastructure.

At each gate, the agent stops and waits. If a file is missing, it surfaces the gap and holds. It does not grind through a blocker. It does not work around the gap silently. It escalates and waits for the operator to decide.

The Sun Run Brief Format

You decide: what project have you been avoiding because it feels like a multi-session grind? That is your first sun run.