temporal-agent-plugin

Provides temporal awareness for AI agents, including timing prediction, progress monitoring, and social timing understanding. Invoke when AI needs time perce...

Security Scan

The skill's code, tests, and SKILL.md are consistent with a library that adds temporal awareness to agents — nothing in the provided material indicates intentional misdirection, but a few development artifacts and network-exposing components deserve manual review before deployment.

Purpose & Capability

The name/description (temporal awareness, progress estimation, social timing) match the included modules (TemporalAnalyzer, ProgressEstimator, SocialTemporal, DistributedClockSync, adapters for LangChain/LlamaIndex) and tests. The listed Python dependencies are plausible for the described data-processing, ML and API functionality.

Instruction Scope

SKILL.md stays on-topic: it documents importing the library, running tests, and starting a local FastAPI server (src/api/endpoints.py). It also shows examples integrating with LangChain/LlamaIndex which naturally require external LLMs. However the SKILL.md (and tests) instruct starting an API server and invoking adapters that will contact external LLM services — you should review src/api/endpoints.py and adapter code to confirm what data is sent, logged, or persisted.

Install Mechanism

There is no automated install spec in the skill bundle (instruction-only), which lowers supply-chain risk. The repo includes requirements.txt and many Python modules; installation would be via pip. No downloads from untrusted URLs were present in the provided files. Manual dependency installation is required and should be done in an isolated environment.

Credentials

The skill declares no required environment variables or primary credential. Adapters reference common LLM clients (ChatOpenAI, OpenAI) which will require user-provided API keys at runtime — that is expected. There are no unexplained secret/env requirements in SKILL.md or the listed files.

Persistence & Privilege

The skill is not forced-always, and does not request elevated agent/platform privileges in metadata. It does include a REST API example (FastAPI), but there is no evidence it modifies other skills or system-wide agent settings. Autonomous invocation is allowed by default (platform normal).

Assessment

This package appears coherent for adding "time awareness" to agents, but before installing or running it: 1) Review src/api/endpoints.py to see what endpoints are exposed and whether it accepts or forwards user data externally; run the API only on localhost or behind an authenticated gateway if you deploy it. 2) Inspect adapter implementations (src/adapters/*) to confirm what they send to external LLMs and whether any logging or telemetry is enabled. 3) Note the development artifact process_base64_zip.py contains hard-coded Windows desktop paths — it can write decoded ZIP contents to disk; review and delete/modify that script if you don't need it. 4) Install dependencies in a virtual environment and run tests locally to validate behavior. 5) Keep your LLM/API keys in a secure secret store and do not embed them in code. If you want, I can scan src/api/endpoints.py and the adapters for network calls, logging, or any code paths that may exfiltrate data (provide their contents) — that would raise my confidence.

智能体时序感知插件

一个为AI及AI框架设计的时序感知插件，解决AI的"体感缺失"问题。

核心功能

1. 任务执行中的"过程体感" - 解决"黑盒焦虑"

• 耗时预估：基于历史执行数据和任务类型，预测完成时间
• 进度感知：提供实时进度更新和预估剩余时间
• 异常检测：识别执行过程中的异常延迟，主动采取措施

2. 用户交互中的"社交体感" - 解决"读不懂空气"

• 动态打断管理：识别用户的停顿和犹豫，合理控制回应时机
• 沉默管理：理解不同长度停顿的语义，避免打断或忽略用户
• 上下文时效性：根据时间窗口判断信息的时效性，提供适当的回应

3. 物理/数字操作中的"因果体感" - 解决"时空错位"

• 动作与反馈时序对齐：预测操作延迟，实现精准控制
• 长流程任务记忆：通过"双时态记忆"保持任务的时间连贯性
• 因果关系预测：理解动作和结果之间的时间延迟，做出合理预判

技术架构

核心组件

• TemporalAnalyzer：时序分析引擎，处理时间相关数据
• ProgressEstimator：进度预估系统，预测任务执行进度和剩余时间
• AnomalyDetector：异常检测模块，识别执行过程中的异常延迟
• SocialTemporal：社交时序理解，理解对话中的停顿和节奏
• CausalPredictor：因果预测模型，预测动作和结果之间的时间关系

框架集成

• LangChainAdapter：与LangChain框架集成
• LlamaIndexAdapter：与LlamaIndex框架集成
• CustomAdapter：与自定义AI框架集成

安装

依赖项

pip install pandas numpy scikit-learn pyarrow fastapi uvicorn pytest python-dateutil scipy matplotlib pytz geopy

使用方法

作为Python库使用

from src.core.temporal_analyzer import TemporalAnalyzer
from src.core.progress_estimator import ProgressEstimator
from src.core.anomaly_detector import AnomalyDetector
from src.core.social_temporal import SocialTemporal
from src.core.causal_predictor import CausalPredictor

# 初始化时序分析引擎
analyzer = TemporalAnalyzer()

# 开始计时
task_id = "download_task"
analyzer.start_timer(task_id)

# 执行任务
# ...

# 停止计时
duration = analyzer.stop_timer(task_id)
print(f"任务执行时间：{duration}秒")

# 预测任务执行时间
predicted_duration = analyzer.predict_duration("analysis", complexity=2.0)
print(f"预测执行时间：{predicted_duration}秒")

与LangChain集成

from src.adapters.langchain_adapter import LangChainAdapter
from langchain.chat_models import ChatOpenAI

# 初始化适配器
adapter = LangChainAdapter()

# 创建语言模型
llm = ChatOpenAI(temperature=0)

# 创建智能体
agent = adapter.create_agent(llm)

# 运行智能体
result = agent.run("帮我分析一份100页的财报，告诉我需要多久")
print(result)

与LlamaIndex集成

from src.adapters.llama_index_adapter import LlamaIndexAdapter
from llama_index.core.llms import OpenAI

# 初始化适配器
adapter = LlamaIndexAdapter()

# 创建语言模型
llm = OpenAI(temperature=0)

# 创建智能体
agent = adapter.create_agent(llm)

# 运行智能体
result = agent.chat("帮我分析一份100页的财报，告诉我需要多久")
print(result)

应用场景

1. AI助手

• 为AI助手添加时间感知能力，让它知道任务需要多长时间
• 当任务执行时间过长时，主动告知用户
• 理解用户的停顿和犹豫，实现自然的对话

2. 自动化工具

• 监控自动化任务的执行进度
• 检测异常延迟，及时采取措施
• 预测任务完成时间，合理安排后续任务

3. 物理操作智能体

• 预测操作延迟，实现精准控制
• 理解动作和结果之间的时间关系
• 保持长流程任务的时间连贯性

技术特点

• 精确的时间感知：基于历史数据和统计分析，提供准确的时间预测
• 实时的进度反馈：实时更新任务进度，提供剩余时间预估
• 智能的异常处理：自动检测异常延迟，主动采取措施
• 自然的社交交互：理解对话中的停顿和节奏，实现自然的交互
• 精准的因果预测：预测动作和结果之间的时间关系，实现精准控制

版本

• 2.0.0：新增贝叶斯预测器、智能超时预测器、分布式时钟同步等功能
• 1.0.0：初始版本，包含所有核心功能

许可

MIT-0