An idle-time, multi-source signal scanner that lets autonomous AI agents generate their own work — bounded by per-action-type cooldowns and role-based availability gates so they become proactive without spamming their own queues.
This repository is a defensive publication. It is published to establish a dated, public, citable record of prior art so that the technique described herein remains free for everyone to practice and cannot be patented by others.
- Publisher / Copyright holder: Gus IT LLC (Florida, USA)
- Author: Gustavo Assuncao, PhD
- Publication date: 2026-06-25
- Version: 1.0
- Document type: Technical Defensive Publication (public prior art)
- Classification: Public
- License: AGPL-3.0-or-later (copyleft; commercial license available)
- Deposit: to be assigned (IP.com / Zenodo / arXiv) — establishes a public, dated, citable prior-art record.
Autonomous AI agents are overwhelmingly reactive: they wait for an inbound trigger — a message, an API call, a webhook — and act only in response. As a result they miss the entire class of proactive opportunities that a competent human handles without being asked: a deadline two hours away, a review that has sat stale for a day, a self-improvement that a learning subsystem already recommended, a teammate who has gone quiet. The naive remedy — "do something every N minutes" — fails in the opposite direction: it spams the agent's own work queue, drowns urgent items, and burns 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 — and each source emits candidate actions. Every candidate is then filtered through two independent gates evaluated before any language-model call is made: (1) an availability gate, a function of the persona's role attribute (e.g. team check-ins are available only to lead personas), 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 (deadline-reminder ≈ 1 h, stale-review ≈ 4 h, self-improve ≈ 12 h), making the agent's "intrinsic motivation" tunable without code change. Only candidates that pass both gates are enqueued. The result is an agent that initiates work like a diligent colleague — bounded, role-aware, budget-respecting — separating self-generated intent from externally-triggered reaction. An illustrative, clean-room reference implementation is included.
We are releasing this as a defensive publication, deliberately and publicly, for one reason: to keep the technique freely practiceable. By placing a complete, enabling, dated technical description into the public domain (under AGPL-3.0-or-later, which carries an express patent grant), we create prior art that bars any party from later obtaining a patent on the core mechanism — the per-action-type layered cooldown, the single-pass multi-source scan, and the pre-LLM role-availability gate. Gus IT LLC chooses openness over exclusivity here: the platform value comes from the integrated product, not from fencing off a scheduling primitive that the whole agent ecosystem benefits from sharing.
| Document | What it contains |
|---|---|
| DEFENSIVE-PUBLICATION.md | The full technical whitepaper: architecture, mechanics, data model, claims, prior-art delta, worked example, framework mapping, evaluation methodology. |
| docs/PRIOR-ART.md | Fuller prior-art landscape, delta table, and honesty attestation. |
| docs/FIGURES.md | All Mermaid figures with captions (Figure 1..N). |
| docs/OPEN-SOURCE-APP.md | Plan for the open-source reference app + Kubernetes/AKS deployment sketch. |
| src/ | Original, illustrative, clean-room reference implementation + src/README.md. |
| LICENSE / NOTICE | AGPL-3.0-or-later license text and attribution. |
| CITATION.cff | Citation metadata. |
This repository is intended to seed a small, self-contained open-source reference application that anyone can run and deploy. See docs/OPEN-SOURCE-APP.md for the planned app, its minimal architecture, how each part maps to the invention, and a generic Kubernetes/AKS deployment sketch.
Licensed under the GNU Affero General Public License v3.0 — see LICENSE. AGPL-3.0-or-later is chosen deliberately for its express patent grant, which reinforces the defensive intent of this publication.
To cite this work, use the metadata in CITATION.cff:
Assuncao, Gustavo. Proactive-Action Engine with Layered Cooldowns — A Technical Defensive Publication. Gus IT LLC, 2026. Version 1.0.
Copyright 2026 Gus IT LLC (Florida, USA). Published 2026-06-25 as a defensive publication / public prior art.