How does the agent decide when to ask for permission?

The skill defines clear boundaries: the agent proceeds with clear requirements but asks for permission for destructive operations, multiple valid approaches, or unclear scopes.

Can this skill help with long-term project handovers?

Yes, by maintaining detailed activity logs with reasoning and 'What's Next' sections, it provides a persistent historical record for both humans and future AI sessions.

Where does this skill store activity logs?

Activity logs are automatically saved to the docs/activity/ directory within your project, organized by date, type, or session for easy retrieval.

Does this include a quality control mechanism?

Yes, it includes a 'Quality Expectations' checklist that ensures code runs, tests pass, and documentation is updated before a task is considered finished.

What is the Agent Behavior skill used for?

It establishes rules for how an AI agent tracks its work, communicates progress, and makes decisions during a coding session to ensure transparency and consistency.

Agent Behavior & Activity Tracking

Name: Agent Behavior & Activity Tracking
Author: bfmcneill

bybfmcneill

0•

生产力与工作流

Establishes standardized activity logging and communication patterns to maintain clear records of AI-driven coding sessions.

This skill optimizes how Claude interacts during development by implementing a structured framework for activity tracking and decision-making. It ensures that every session is documented within the project's own repository, providing a clear audit trail of changes, technical decisions, and trade-offs. By defining specific protocols for when to ask questions versus when to proceed independently, and establishing rules for subagent usage, it streamlines the workflow, reduces ambiguity, and maintains high quality through a built-in 'Definition of Done' checklist.

主要功能

01Defined 'Ask vs. Proceed' protocols to minimize interruptions while maintaining safety

02Optimized subagent management for parallelizable and deep exploration tasks

030 GitHub stars

04Integrated quality assurance checklist for final task verification

05Structured session templates covering changes, reasoning, and next steps

06Standardized activity logging in docs/activity/ with hierarchical file naming

使用场景

01Onboarding an AI agent to a new codebase with clear communication boundaries

02Managing complex, multi-session features where state tracking and documentation are critical

03Maintaining a transparent audit trail for team-based AI coding projects

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bfmcneill/agi-marketplace agent-behavior

For use in Claude.ai and ChatGPT

Download Skill