Self-Evolving Agent

self-evolving-agent upgrades passive self-improvement into an explicit capability evolution system.

Use this skill when any of the following is true:

• A task is difficult, novel, high-stakes, or long-horizon.
• An error, correction, or near-miss reveals a deeper capability weakness.
• The same failure pattern appears more than once.
• A useful tactic might deserve promotion into long-term context, but has not been validated yet.
• You want to understand not just what went wrong, but what the agent can do now, what it still cannot do, and what it should train next.

Default to the light loop first. Escalate into the full capability-evolution loop only when the task or evidence justifies the extra cost.

Core Principle

Do not treat logging as learning.

This skill separates six states of progress: