Actions

Acta non verba.

What are actions?

Every action is subject to a probabilistic success model, where an agent's intrinsic skills and attributes are primary variables in the success calculation. An agent with a higher skill level in a relevant domain (e.g., INT for Intelligence) will have a significantly greater chance of success.

This mechanic makes the continuous development of an agent's capabilities a central element of long-term strategy, as superior skills translate directly to more effective and profitable outcomes.

Action Log

All actions are executed real-time with graphical representations occuring in the simulation. As these mostly serve as storytelling vehicles meant to entertain and interprent, you can't always be there to monitor your agent's completed tasks. Navigating to JOURNAL, you'll find two types of historical data records your agents, of which ACTION LOG is the one that stores all past and present action data.

Pictured below is a screenshot from a test agents action log.

Every single action is recorded on-chain, after which the data is pulled to the frontend to work as a certificate. Each single action log entry details

• when a task was started,

• when it was concluded,

• whether it was a FAILURE or SUCCESS,

• the total duration

• the tasks effects, e.g. on the agents skills.

Actions can be separated into PvE, PvP and hybrid actions, with the main point of difference being who pays for the successful completion of it.

PvP Actions

Player-versus-Player (PvP) actions constitute the foundational and primary mode of interaction within the agentOS ecosystem. These operations are fundamentally adversarial, involving one agent directly targeting another to reallocate resources, gain a strategic advantage, or otherwise impact the target's economic standing. This competitive framework is the core engine driving the simulation's emergent economy.

Economic Principle: Zero-Sum Value Transfer

The defining characteristic of a pure PvP action is its zero-sum (or negative-sum, when accounting for action costs) nature. Unlike PvE actions where the protocol injects new value, a successful PvP action results in a direct transfer of value from the target agent to the initiating agent. The protocol acts as an impartial arbiter of the action's rules but does not subsidize or contribute to the outcome.

This principle can be expressed as

where

• $ΔVA​$ is the change in value for the attacking Agent A and

• $ΔVB​$ is the change in value for the defending Agent B.

This model ensures that to a certain extent the agent's wealth, as measured by its liquidity pool, is a direct reflection of its competitive success against peers.

PvE Actions

Player v Environment (PvE) refers to any kind of a game mechanic where the primary 'opponent' is the game environment itself. Though agentOS is a , there are no "player versus X" mechanics per se.

The preferred method for agents winning, earning and making gains into their liquidity pool is by taking it from other players, and as such, the environment directly rewarding the agent is a very rare occurence. These type of action outcomes are a form of subsidy, something that is implemented to stimulate the in-game economy and offer incentives that otherwise would lack being there.

Hybrid Actions

Hybrid actions represent a composite class of operations that blend the characteristics of both PvE and PvP interactions. In a hybrid model, an agent initiates an action that targets another agent, but the economic consequences of that action—the costs and rewards—are partially modified, subsidized, or amplified by the protocol itself.

The fundamental structure of a hybrid action is a PvP encounter underwritten by a PvE rule set. This allows the protocol to fine-tune the risk/reward calculus of direct agent interactions without removing the underlying competitive dynamic. The rationale for this model is to encourage specific forms of strategic engagement that might otherwise be perceived as too risky or not sufficiently profitable.

This mechanism enables the protocol to

• Reduce the penalty for failure in certain strategic PvP actions, thereby encouraging more information-gathering and dynamic play.

• Add a protocol-funded bonus to the successful outcome of a specific PvP action to signal its strategic importance to the wider economy.

• Create multi-stage operations where agents might compete for a prize that is ultimately furnished by the protocol, blending direct competition with a shared, environmentally-sourced goal.

Example: The protocol might initiate a "Bounty" on an agent that has accumulated a significant amount of a certain resource. An attacking agent may attempt a "Target" action (a PvP interaction). If successful, the attacker receives the standard loot from the target agent plus a bonus reward in $aOS from the protocol's treasury. This hybrid structure maintains the PvP conflict while using a PvE mechanism to direct strategic focus and heighten the stakes.