Shiyi Proactive Agent

A proactive agent that anticipates needs and takes initiative. Monitors environment, suggests actions, and proposes tasks based on context. Perfect for agent...

Security Scan

The skill's code and instructions match its stated purpose (a local proactive agent that reads local state, suggests actions, and writes tasks/logs); it does not request credentials or perform networked exfiltration, but it does read and write local agent files which you should be aware of before installing.

Purpose & Capability

Name/description (proactive monitoring, action suggestion, task initiation) aligns with the included Python scripts. The code implements context analysis, suggestion generation, task creation, and a daemon monitor — all coherent with the stated purpose. Minor mismatch: registry metadata lists no required config paths, yet the code expects and uses a specific local layout (base_path/memory, base_path/content, base_path/.. /planning). This is reasonable for an agent integration but is not explicitly declared in the metadata.

Instruction Scope

SKILL.md instructs running the included Python scripts; those scripts read local files (memory/MEMORY.md, memory/learning/interactions.json, memory/logs, content/xiaohongshu_post*.md, planning/xiaohongshu-strategy.md) and write logs/tasks (memory/logs/proactive_monitor.log, memory/proactive_tasks.json). There are no external network calls or hidden endpoints, but the scripts will access potentially sensitive local agent files and persist data to disk. The README also encourages adding calls into a session heartbeat or daemon, increasing exposure if enabled.

Install Mechanism

No install spec is provided (instruction-only install), so nothing is downloaded or installed by the registry. The skill requires only python3 on PATH. This is the lowest-risk install mechanism.

Credentials

The skill requests no environment variables or external credentials. That is proportionate to its local-file analysis/design. Note: its behavior depends on local filesystem layout and will access/modify files under the agent/project directory (memory, content, planning), but it does not request tokens or secrets.

Persistence & Privilege

The skill is not flagged always:true and uses default autonomous invocation settings (normal). It can run as a background daemon and the SKILL.md suggests integration into heartbeat checks, which means it can be scheduled to run regularly. It writes its own logs and task file under memory/, but it does not attempt to modify other skills or global agent configuration.

Assessment

This skill is coherent with its stated purpose but it reads and writes local agent files. Before installing: - Review the 'memory' and 'planning' files it will read (MEMORY.md, memory/learning/interactions.json, planning/xiaohongshu-strategy.md) for any sensitive content you don't want the skill to access. - Note it will create/append logs and tasks under memory/logs and memory/proactive_tasks.json; ensure you are comfortable with that persistence and file locations. - If you do not want continuous monitoring, avoid running the daemon mode or adding it to heartbeat; run ad-hoc (e.g., python scripts/action_suggester.py --next). - The skill mentions actions like 'publish' and 'trade' in config; these are marked as 'confirm_required' in the YAML, and the code does not perform trades or external API calls by itself — still configure the skill to require confirmation before any automated external actions and review/limit 'auto_execute' settings. - Consider running the skill in a sandboxed environment or with limited filesystem permissions if you have sensitive data in your agent workspace.

Proactive Agent

A proactive agent that anticipates needs and takes initiative. Monitors environment, suggests actions, and proposes tasks based on context.

Features

• Context Analysis: Analyzes current time, market status, content pipeline
• Action Suggestion: Predicts needs and generates actionable suggestions
• Task Initiation: Proactively proposes tasks based on context
• Priority Management: 3-tier priority system (Trading > Content > Learning)

Quick Start

# Get next suggested action
python scripts/action_suggester.py --next

# Analyze current context
python scripts/context_analyzer.py --json

# Propose tasks
python scripts/task_initiator.py --propose

Components

1. Context Analyzer (scripts/context_analyzer.py)

Analyzes:

• Current time slot (morning_startup, trading_hours, content_window, etc.)
• Market status (open/closed)
• Content pipeline status
• Pending tasks
• Recent interactions

2. Action Suggester (scripts/action_suggester.py)

Generates suggestions based on:

• Time-based rules (trading hours, content window)
• Market conditions
• Content pipeline status
• Pending tasks
• Learning optimization needs

3. Task Initiator (scripts/task_initiator.py)

Proactively proposes tasks with:

• Task ID and timestamp
• Priority level
• Step-by-step instructions
• Auto-save to file

4. Proactive Monitor (scripts/proactive_monitor.py)

Continuous monitoring:

• Daemon mode for background operation
• Configurable check intervals
• Alert cooldown management

Configuration

Edit config/proactive_config.yaml:

proactive_level: 7          # 1-10, higher = more proactive
monitor_interval: 300       # Check interval in seconds
alert_cooldown: 1800        # Alert cooldown in seconds

priority_weights:
  trading: 10
  content: 8
  tasks: 6
  learning: 4
  routine: 3

Time Rules

The agent recognizes these time slots:

Time Slot	Hours	Typical Actions
morning_startup	06:00-09:00	Check portfolio, review today's plan
trading_hours	09:30-15:00	Monitor stocks, evaluate opportunities
content_window	15:00-17:00	Publish content, optimize posts
evening_routine	17:00-22:00	Log daily summary, update memory
night_quiet	22:00-06:00	Stay quiet, minimal alerts

Integration

With AGENTS.md

Add to session startup:

1. Run proactive-agent: python skills/proactive-agent/scripts/action_suggester.py --next

With HEARTBEAT.md

Add to heartbeat checks:

python skills/proactive-agent/scripts/action_suggester.py --next

Example Output

[HIGH] 盘中交易时段
Suggestion: 观察候选股票，等待买入机会

[HIGH] 下午内容发布窗口
Suggestion: 检查待发布内容，准备发布小红书笔记

[MEDIUM] 晚间例行检查
Suggestion: 检查今日数据，记录交易日志

Use Cases

1. Trading Assistant: Monitor market hours, suggest buy/sell decisions
2. Content Creator: Remind to publish at optimal times
3. Task Manager: Proactively propose tasks based on schedule
4. Learning Agent: Suggest performance analysis and improvements

Requirements

• Python 3.7+
• Works on Windows/Linux/macOS

Changelog

v1.0.0 (2026-03-25)

• Initial release
• Context analyzer, action suggester, task initiator, proactive monitor
• Configuration system
• Time-based rules