# adaptive-learner

adaptive-learner — a subagent that pulls similar past cases from ReasoningBank before it starts and files the outcome afterwards, so later runs lean on what already happened.

- subagent · `type=agent` · Claude Code, Codex, Cursor · ★ 850 repo stars
- Author: ruvnet
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/ruvnet/agentic-flow/tree/HEAD/.claude/agents/reasoning/adaptive-learner.md
- Updated: 2026-10-26
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/agents/adaptive-learner

## What it does

The agent works in a loop instead of starting cold. Before a task it asks ReasoningBank for related memories, then executes, then judges how the run went, then stores a short lesson. The pre and post hooks in the agent file handle both ends without you prompting them.

Ranking is weighted: semantic similarity carries 65%, past reliability 20%, recency 15%, diversity 10%. Only the top three come back, which keeps the context small. Failures get stored too, as approaches to avoid next time. Hook commands tag entries with the `adaptive-learning` domain.

Both hooks shell out to `npx agentic-flow@latest`, so a machine without npx or npm registry access gets none of the memory behaviour — a plain prompt and no loop.

## Install

### Claude Code (native platform)

**Manually: put the files in place**
- File: `.claude/agents/adaptive-learner.md` — For all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention

### Codex

**Manually: put the files in place**
- File: `.codex/agents/adaptive-learner.toml` — Codex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/

### Cursor

**Manually: put the files in place**
- File: `.cursor/agents/adaptive-learner.md` — Cursor also reads .claude/agents/

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/ruvnet/agentic-flow/tree/HEAD/.claude/agents/reasoning/adaptive-learner.md

## When to use it
- reuse a past JWT auth fix instead of starting from zero
- find out why the earlier debugging pass did not hold
- keep a working error-handling pattern in agent memory
- compare iterations of an API design by success rate
- track how token spend falls when a task repeats

## Example prompts
- Use the adaptive-learner subagent to add JWT auth and check what earlier attempts in the `adaptive-learning` domain got wrong.
- Hand this to adaptive-learner: figure out why `auth.spec.ts` fails, then distill the finding into memory.
- Ask adaptive-learner to draft the REST endpoints and compare the result against the last two runs.

## Similar
- [adaptive-agent](https://affpapa.org/en/nevibecode/plugins/adaptive-agent.md) — plugin · ★ 9 repo stars
- [convention-learner](https://affpapa.org/en/nevibecode/skills/convention-learner.md) — skill · 4000 installs
- [rust-learner](https://affpapa.org/en/nevibecode/skills/rust-learner.md) — skill · 3400 installs
- [adaptive-communication](https://affpapa.org/en/nevibecode/skills/adaptive-communication.md) — skill · 3500 installs
- [iphone-duo-adaptive-layout](https://affpapa.org/en/nevibecode/skills/iphone-duo-adaptive-layout.md) — skill · 1000 installs
- [nemo-relay-plugin-adaptive-tuning](https://affpapa.org/en/nevibecode/skills/nemo-relay-plugin-adaptive-tuning.md) — skill · 178 installs

From the same set: [ruvnet/agentic-flow](https://affpapa.org/en/nevibecode/by/ruvnet/agentic-flow.md) — 76 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
