hz-proactive-agent
by lidekahdjdhdhsjjs-lang
Productivity & TasksBrief
Transform AI agents from task-followers into proactive partners that anticipate needs and continuously improve.
Description
What This Skill Does
A behavioral architecture for AI agents that adds proactive action, memory persistence, and self-improvement. Defines protocols for capturing corrections before responding, surviving context truncation, running background tasks autonomously, and evolving safely over time.
Standard agents lose state between sessions and wait for instructions; this provides concrete file-based protocols and trigger rules that make persistence and proactive behavior reliable by default.
When to Use It
- Automatically saving a user correction like a color preference before the agent responds
- Recovering full context after a long session gets compacted or truncated
- Running memory maintenance as an isolated background cron without interrupting active work
- Surfacing unprompted suggestions based on patterns in past conversations
- Auditing all config files and cron prompts when migrating away from a deprecated tool
Example Workflow
User says: 'Use the blue theme, not red'
- 1
Detects a correction trigger in the message
- 2
Stops before composing a response
- 3
Writes 'Theme: blue (not red)' to SESSION-STATE.md
- 4
Confirms the write succeeded
- 5
Responds to the user
Correction is durably stored in SESSION-STATE.md and survives context compaction
Installation
clawhub install lidekahdjdhdhsjjs-lang/hz-proactive-agentRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.