skills.rest
Wike-CHI
Wike-CHICOMMUNITY
Explore all skills by this author →
WebsiteGitHub

proactive-agent

Persist session-critical decisions and context to SESSION-STATE.md before responding.

2|Updated Mar 31, 2026
Search Tags:#self-improvement#agent#memory#proactive#wal#working-buffer#compaction-recovery
One-click install
npx skills add https://github.com/Wike-CHI/acquisition-agent --skill proactive-agent-wike-chi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: proactive-agent
Source: https://github.com/Wike-CHI/acquisition-agent/tree/main/skills/proactive-agent
Command: npx skills add https://github.com/Wike-CHI/acquisition-agent --skill proactive-agent-wike-chi

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Proactive-agent prevents AI assistants from losing critical context, failing background tasks, and repeating work when sessions are truncated or memory is compacted. It ensures decisions, corrections, and important details are captured reliably so the agent can continue work across restarts and run autonomous maintenance without human prompts.

Core Features & Use Cases

  • WAL Protocol (Write-Ahead Log): Persist corrections, decisions, proper nouns, and specific values to SESSION-STATE.md before responding.
  • Working Buffer: Capture every exchange in the "danger zone" after context usage passes a threshold so no recent interaction is lost during compaction.
  • Compaction Recovery: Read the working buffer and session state to reconstruct context and resume interrupted tasks without asking the user "what were we doing?"
  • Autonomous vs Prompted Crons: Use isolated agentTurns for background work and systemEvents for attention-requiring prompts to avoid missed maintenance tasks.
  • Self-Improvement Guardrails & Verification: Safe evolution via ADL/VFM rules, verify implementation changes (not just text), and enforce security hardening before installing skills or acting externally.
  • Use Cases: Auto-populate onboarding files and USER.md, run periodic memory fresheners and heartbeat audits, recover a project after a context truncation, and generate draft proactive suggestions without sending externally.

Quick Start

Ask the agent: "Before replying, write any corrections, decisions, or proper nouns into SESSION-STATE.md, then summarize the current task and next steps."

Frequently Asked Questions about proactive-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent AI agent memory loss during context compaction?▼

To prevent memory loss during context compaction, use a write-ahead logging protocol to persist corrections, decisions, and proper nouns into a dedicated session state file before the agent responds. A working buffer captures exchanges in the danger zone past context thresholds.

Can I run autonomous background crons without missing maintenance tasks?▼

Yes, you can run autonomous background crons by using isolated agentTurns for background work and systemEvents for prompts requiring attention. This separation ensures autonomous maintenance tasks execute without missing critical human-in-the-loop interventions.

What are the security guardrails for proactive agent self-improvement?▼

Proactive agent self-improvement requires security guardrails enforcing ADL/VFM rules for safe evolution, verifying implementation changes rather than just text, and applying strict security hardening before installing skills or executing any external actions.

Why does my long-running conversational agent repeat work after a restart?▼

Long-running conversational agents repeat work after restarts because critical context was not captured to a persistent session state file. Applying a write-ahead log ensures decisions are saved before responding, enabling seamless continuity across restarts.

Related Skills

Explore More Software Engineering Skills→
bluegitterbluegitter
1

proactive-agent

Preserve session-critical details in SESSION-STATE.md with write-ahead logging and compaction recovery.

Community
Advanced
papachongpapachong
2

proactive-agent

Persist agent decisions and context using write-ahead logging before responding.

Community
Advanced
PandaWithAPlanPandaWithAPlan
2

psao-annotate

Annotate prompts with metadata tags to prevent semantic drift during task delegation.

Community
Intermediate
MerkyorLynnMerkyorLynn
37

proactive-agent

Implement proactive agent operations with session-state persistence and recovery.

Community
Advanced