| name | ax-java-agent-memory-skills |
| description | Use when writing Java code with `dev.axllm:ax` for agent memory, recall callbacks, dynamic skill discovery, loaded-skill state, and used-skill tracking. |
| version | 22.0.6 |
AxAgent Memory And Skills For Java
This skill helps an agent write Java code with the generated Ax package dev.axllm:ax. Use the generated package API, examples, and manifests; do not import TypeScript-only APIs unless you are editing the TypeScript package.
When To Use
- Load memories or skill guides into an RLM agent run.
- Track which memories or skills actually influenced a turn.
- Keep recall and skill search as host callbacks rather than generated-package global state.
Package Facts
- Language: Java.
- Package:
dev.axllm:ax.
- Package API docs:
API.md and axir-api.json.
- Capability manifest:
axir-capabilities.json.
- Runnable examples:
examples/.
- Real network support: yes.
- Scripted no-key transport support: yes.
- Runtime profiles:
javascript-quickjs, python-pyodide.
Core Pattern
AxAgent helper = Ax.agent("question:string -> answer:string", java.util.Map.of());
var out = helper.forward(llm, java.util.Map.of("question", "How should I proceed?"));
Relevant API Surface
- Agents And RLM:
Ax.agent, AxAgent
- Runtime Profiles:
ProcessCodeRuntime, RuntimeCapabilities, RuntimeEnvelope, javascript-quickjs, python-pyodide
Guardrails
- Start from package examples for exact native syntax before inventing a new call shape.
- Use
provider-api examples only when the user explicitly has provider credentials available.
- Use
no-key examples for deterministic local checks and provider request mapping.
- Treat AxIR as the source of generated package truth: if package docs disagree with source code, update the compiler and regenerate packages.
- Do not copy repo-maintainer skills from
tools/*/skills/ into user packages.
