Skip to main content

ax-cpp-playbook

Use when writing C++ code with `axllm` for the playbook() context-engineering surface, agent-bound verified evolution, run-end learning, online updates, and rendering a playbook into a program.

Informations de source

Dépôt
ax-llm/ax
Dernière activité de la source
15 octobre 2026 à 09:33
Langue détectée de SKILL.md
anglais
Étoiles
2 967
Forks
198

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

Affichage de SKILL.md

SKILL.md
Instructions source · Aperçu en lecture seule
name
ax-cpp-playbook
description
Use when writing C++ code with `axllm` for the playbook() context-engineering surface, agent-bound verified evolution, run-end learning, online updates, and rendering a playbook into a program.
version
25.4.1
# Ax Playbook For C++ This skill helps an agent write C++ code with the generated Ax package `axllm`. Use the generated package API, examples, and manifests; do not import TypeScript-only APIs unless you are editing the TypeScript package. ## When To Use - Grow an evolving context playbook for a program or agent stage with playbook(). - Attach a seed playbook to an agent and learn bounded avoidance rules from run-end failure signals. - Use the agent-bound playbook evolve method to mine grounded weaknesses with verification and exact rollback. - Refine a playbook online from live feedback or offline from labeled examples. - Render or persist a playbook and inject it into a program context. ## Package Facts - Language: C++. - Package: `axllm`. - 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 ```cpp auto program = axllm::ax("question:string -> answer:string"); auto pb = axllm::playbook(program, *llm); pb.evolve(examples, metric_fn); ``` ## Relevant API Surface - Optimizers: `axllm::optimize`, `axllm::playbook`, `axllm::AxPlaybook`, `axllm::AxBootstrapFewShot`, `axllm::AxGEPA`, `axllm::OptimizerEngine`, `axllm::OptimizerEvaluator` ## 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.
Voir sur GitHub