Results for “recursive-self-improvement”

Implement iterative evaluation and refinement loops for AI agent outputs, using self-critique, evaluator-optimizer patterns, and rubric-based scoring to improve quality.

Autonomously optimize any Claude Code skill by running it repeatedly, scoring outputs against binary evals, mutating the prompt, and keeping improvements. Based on Karpathy's autoresearch methodology. Use when: optimize this skill, improve this skill, run autoresearch on, make this skill better, self-improve skill, benchmark skill, eval my skill, run evals on. Outputs: an improved SKILL.md, a results log, and a changelog of every mutation tried.

Evaluates the agent's own work, catches mistakes, and improves permanently through self-reflection, self-criticism, and learning from corrections.

Run autonomous optimization loops that iteratively improve artifacts against evaluators using hypothesis tree refinement, without overfitting.

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training

Analyzes repeated agent failures or noisy routing and proposes evidence-backed changes to instructions, skills, gates, or runbooks.

Run Karpathy-style autonomous ML search on a real training repo: choose the right mode (setup, program.md, bounded loop, results interpretation, or constrained-hardware adaptation), preserve the immutable prepare.py / 300-second / val_bpb contract, and route prompt/skill eval work away to LangSmith, Promptfoo, Braintrust, or skill-autoresearch.

自动捕获经验并安全进化的技能。触发条件：(1)命令/操作失败时→记ERRORS.md (2)被用户纠正('不对'/'应该是')时→记LEARNINGS.md (3)用户需要不存在的能力时→记FEATURE_REQUESTS.md (4)外部API/工具出错时→记ERRORS.md (5)发现自己知识过时/错误时→记LEARNINGS.md (6)发现更好做法时→记LEARNINGS.md (7)每个任务完成时→回顾过程，有新经验则记LEARNINGS.md。去重原则：如果没有新经验或已有条目已覆盖则跳过不写。每次写入同时在.learnings/CHANGELOG.md追加JSONL日志。经验反复出现≥3次时晋升到AGENTS.md/TOOLS.md/SOUL.md。详见正文。

Expert guidance for implementing GRPO/RL fine-tuning with TRL for reasoning and task-specific model training.

Turn owner feedback about agent behavior into concrete system changes. Use when the owner says something is off, wants the assistant to improve how it operates, asks for a reflection, or wants a durable fix instead of a one-off apology.

Runs an autonomous optimization loop that iteratively improves an artifact against an objective and evaluator using Hypothesis Tree Refinement, with subagent executors in isolated git worktrees.

Iteratively improve a Claude Code skill using the skill-reviewer agent until it meets quality standards. Use when improving a skill with multiple quality issues, iterating on a new skill until it meets standards, or automated fix-review cycles instead of manual editing.

Patterns and techniques for evaluating and improving AI agent outputs. Use this skill when: - Implementing self-critique and reflection loops - Building evaluator-optimizer pipelines for quality-critical generation - Creating test-driven code refinement workflows - Designing rubric-based or LLM-as-judge evaluation systems - Adding iterative improvement to agent outputs (code, reports, analysis) - Measuring and improving agent response quality

Continuously monitors system performance, identifies improvement opportunities, and orchestrates skill discovery and creation to autonomously evolve capabilities.

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training

Builds reliable browser automation skills by iteratively running a browsing task, reading the trace, and improving the navigation strategy until it passes consistently.

Expert guidance for GRPO/RL fine-tuning with TRL for reasoning and task-specific model training