agent-time-estimation

Recalibrate time and effort estimates for engineering and coding tasks to reflect AI-agent execution speed instead of human-developer throughput. Consult this skill any time you are about to state how long an engineering or coding task will take — whenever you give an ETA, size a change, commit to a delivery, or emit any task/effort duration ("this will take 1-2 days", "a couple hours", "about a sprint", "ready by end of week"). Claude systematically overestimates engineering time by one to three orders of magnitude because its priors are calibrated on human developer productivity, not on how fast an agent actually executes — so use this even when the user did NOT explicitly ask "how long will this take". Also triggers on planning, roadmapping, sprint sizing, story points, effort/complexity estimation, deadlines, and phrases like "how long", "ETA", "when will this be done", "by [date]", "how much work is this". Scope: engineering/coding *execution* time — not a user's own manual effort, calendar scheduling, or non-technical durations.