How does the North Star anchoring help?

North Star anchoring establishes a vision, success criteria, and a 'do-not-do' list at the beginning of a task, ensuring the agent stays focused on the core objective and avoids scope creep.

Is it safe to let the agent modify my code autonomously?

Yes. The skill uses Worktree isolation to keep changes in a separate environment and includes mandatory checkpoints (CP0-CP6.5) that enforce architectural rules and testing before any task is considered complete.

What is a Self-Evolving Agent in Claude Code?

It is a specialized skill that allows Claude to autonomously iterate on a specific goal by learning from its environment, checking its progress against a 'North Star' vision, and refining its strategy until success is achieved.

Can I control how many times the agent iterates?

Absolutely. You can use the --max-iterations flag (defaulting to 10) to limit how many PDCA cycles the agent performs before pausing for human intervention.

Self-Evolving Agent

Name: Self-Evolving Agent
Author: miles990

bymiles990

生産性とワークフロー

Enables autonomous goal achievement through iterative learning, structured PDCA cycles, and self-improving workflows.

概要

The Self-Evolving Agent skill transforms Claude into a highly disciplined, goal-oriented agent capable of independent problem-solving. It utilizes a structured 'North Star' anchoring system to maintain focus, Worktree isolation for safe experimentation, and a persistent Git-based memory to learn from every iteration. By following a rigorous Plan-Do-Check-Act (PDCA) loop and navigating mandatory architectural checkpoints, this skill is ideal for complex engineering tasks that require multi-step reasoning, systematic debugging, and continuous refinement without constant human oversight.

主な機能

Autonomous PDCA Cycle for iterative task execution and self-correction.
Git-based Memory System that records lessons learned for future task optimization.
Mandatory Guardrail Checkpoints (CP0-CP6.5) ensuring architectural consistency and code quality.
North Star Anchoring to prevent scope drift and define clear success criteria.
Worktree Isolation for managing parallel tasks in a safe, sandboxed environment.
0 GitHub stars

ユースケース

Executing complex codebase refactors or migrations with continuous TDD verification.
Systematic performance optimization and debugging using multi-strategy retry logic.
Developing end-to-end features from high-level specifications using autonomous learning.

概要

主な機能

Autonomous PDCA Cycle for iterative task execution and self-correction.
Git-based Memory System that records lessons learned for future task optimization.
Mandatory Guardrail Checkpoints (CP0-CP6.5) ensuring architectural consistency and code quality.
North Star Anchoring to prevent scope drift and define clear success criteria.
Worktree Isolation for managing parallel tasks in a safe, sandboxed environment.
0 GitHub stars

ユースケース

Executing complex codebase refactors or migrations with continuous TDD verification.
Systematic performance optimization and debugging using multi-strategy retry logic.
Developing end-to-end features from high-level specifications using autonomous learning.