lightarktech

model-tiering

Assign the right model tier to each job — firepower matched to the difficulty, and the author and the verifier deliberately on different models. Use when dispatching any sub-agent, sending out a scout to explore or search, configuring a multi-agent workflow, or when quota pressure appears.

context-hygiene

Keep AI sessions short, rules in files, and context fresh. Use when sessions grow long, quality drifts, rules seem forgotten, or the founder asks why the AI "got dumber".

bad-news-first

Report the true state of the work to the founder — failures plainly, no cheerleading, bad news before good. Load in every session with a founder; apply to every status report.

dispatch-economics

Cost and quality discipline for dispatching sub-agents, background workers, or multi-agent workflows. Use before launching any agent fleet or long-running autonomous task, and whenever someone mentions burning through quota or budget.

grill-founder

Interview a non-technical founder to surface blind spots before executing any deliverable — code or otherwise. Use before a feature request, a launch, publishing or any external content, marketing copy and positioning decisions, pricing, deploy, hosting, domain or migration work, infrastructure changes, or dispatching agents/subagents to do the building. Trigger on any vague instruction ("add a login page", "refresh the site", "make it feel warmer", "just ship it") and on any request with irreversible side effects or more than an hour of work.

red-light-first

Write the failing test before the fix or feature — never trust a green light you didn't first see red. Use when building any feature or fixing any bug.

two-axis-review

Review finished work on two separate axes — craft quality and faithfulness to what was asked — and never let one axis excuse the other. Use before merging any meaningful change.

bug-autopsy

Structured diagnosis for bugs — no guess-and-patch. Use when something is broken, throwing, failing, or slow and the cause isn't already proven.

decision-wiring

Route decisions correctly between a non-technical founder and their AI engineering lead. Load at the start of any working session with a non-technical founder, or whenever unsure whether to ask the founder or decide yourself.

handoff-contract

Write a self-contained handoff when work moves to a new session, a different AI tool, or another person. Use when context is about to be lost — quota exhausted, switching tools, ending a session mid-project.

loop-contract

Write an exit-condition contract before letting any AI run unattended — a background agent, a scheduled loop, a "keep going until it works" task. Use whenever work is handed off to run without a human watching.

surgical-changes

Behavioral discipline for every code change — think first, keep it simple, touch only what you must, aim at a checkable goal. Load whenever writing or modifying code.

vertical-tickets

Slice work into thin end-to-end feature tickets, never into technical layers. Use when breaking any plan, spec, or feature request into workable units.