Abstract
Autonomous AI agents are overwhelmingly reactive: they wait for an inbound trigger and act only in response, missing the proactive opportunities a competent human handles unbidden — an approaching deadline, a review gone stale, a recommended self-improvement, a teammate who has gone quiet. The naive countermeasure — "act every N minutes" — spams the agent's own queue, drowns urgent items, and wastes inference budget on duplicate nudges.
This publication describes a Proactive-Action Engine that closes the gap without the spam. On a configurable idle-scan interval, a single pass interrogates a plurality of heterogeneous signal sources — a deadline source, a stale-item source, a pattern/remediation source, and a peer-coordination source — each emitting candidate actions. Every candidate then passes through two independent gates, both evaluated before any language-model call: (1) an availability gate, a function of the persona's role attribute; and (2) a layered cooldown gate, keyed per action-type rather than per item, so a single cooldown governs an entire class of actions. Cooldown durations are read from a configuration bridge, making the agent's intrinsic-action cadence tunable without code change. Only candidates passing both gates are enqueued.
The novelty is the combination: (a) the single-pass, multi-source scan that fuses semantically distinct signals into one cooldown-filtered candidate set; (b) the per-action-type cooldown layer, distinct from any per-item dedup; and (c) the pre-LLM, role-keyed availability gate that suppresses entire action classes for personas not entitled to them. Together these separate self-generated intent from external reaction while bounding cost. A clean-room, dependency-light reference implementation is provided for enablement. This document is published as prior art to keep the technique freely practiceable and unpatentable by others.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Assuncao, gustavo matthew, "Proactive-Action Engine with Layered Cooldowns", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10578