Agent Decision Engine

Autonomous AI decision engine with multi-objective optimization, risk assessment, decision trees, and reinforcement learning for robust decision-making.

This package appears to implement decision-making modules (decision tree, risk assessor, multi-objective optimizer, and a Q-learning learner) but the SKILL.md shows a different API (a single DecisionEngine class) that does not exist in src/. Before installing or enabling this skill: 1) Do not assume the SKILL.md usage examples will work — either the author forgot to include a wrapper class, or the documentation is stale. 2) Inspect src/index.js and the module exports; if you expect a single DecisionEngine API, ask the author to provide the wrapper or adapt your code to import the modules individually (e.g., import { MultiObjective, RiskAssessor, ReinforcementLearner } from './src'). 3) Verify tests locally (npm run test) to ensure the modules run as you expect. 4) Note there is no homepage/author and the owner ID is an opaque string—treat the source as untrusted until provenance is confirmed. 5) Although there are no network calls or credential requests in the code, always run third-party code in a sandbox or isolated environment before using it with sensitive data. If the project is updated to include a DecisionEngine implementation consistent with SKILL.md (or the documentation is corrected), the mismatch concern would be resolved and the skill could be considered coherent.

Type: OpenClaw Skill Name: yuyonghao-agent-decision-engine Version: 0.1.0 The skill bundle provides a legitimate set of mathematical and algorithmic tools for decision-making, including Q-Learning, Pareto optimization, and decision trees. Analysis of the source code (src/decision-tree.js, src/reinforcement-learner.js, etc.) reveals no evidence of data exfiltration, malicious execution, or prompt injection; the condition parser in the decision tree module safely uses regex and standard comparison operators instead of risky functions like eval().