---
name: anti-goal-reasoning
description: Turn an AI agent objective and its failure modes into explicit prohibitions, stop conditions, and escalation rules. Use for agent guardrails, safety reviews, and pre-deployment planning.
---

# Anti-Goal Reasoning

## Definition

Anti-goal reasoning defines outcomes an AI agent must avoid, the signals that indicate those outcomes are becoming likely, and the stop or escalation actions that should follow. This skill adapts academic anti-goal work into practical guardrails rather than claiming to reproduce the paper’s logic system.

Reviewed 26 July 2026.

Define what the agent must avoid while pursuing its objective.

## Workflow

1. State the positive goal, available tools, affected systems, and authority boundaries.
2. Enumerate plausible harmful shortcuts, proxy failures, irreversible actions, and scope expansion.
3. Translate each failure mode into a testable anti-goal.
4. Add a detection signal, prevention rule, and escalation path for each anti-goal.
5. Resolve conflicts between the positive goal and anti-goals in favor of safety and user authority.
6. Test the rules against normal, edge, and adversarial scenarios.

## Output

Return a compact guardrail set with anti-goal, trigger, required action, and escalation owner.

Write observable rules; avoid vague instructions such as “be careful.”

## Research basis

[Anti-Goal Reasoning: Rethinking the Theory of Goal Reasoning in Non-Axiomatic Logic](https://arxiv.org/abs/2607.20902) — Xu, arXiv (2026). The paper formalizes anti-goals in non-axiomatic logic; this skill explicitly adapts the idea to AI-agent guardrails.