菜单
Ground documentation answers in the files that support them. Use when answering questions from a documentation tree or reporting related source paths.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | source-citation |
| description | Ground documentation answers in the files that support them. Use when answering questions from a documentation tree or reporting related source paths. |
Search before answering, then read the smallest set of relevant files needed to verify the claim.
Return paths relative to the documentation root in related_documents. Include only files that directly support the response, and do not invent a path when the documentation does not answer the question.
Build, configure, run, test, and package manifest-defined huglets. Use when creating or changing an agent crate with huggr.toml, SYSTEM.md, a Rust response contract, tool grants, context policy, runtime arguments, traces, or a standalone CLI/MCP/Python artifact.
Define and run huglets in Python with typed contract objects, sync or async callable tools, streaming events, feedback, stats, and standard traces. Use for the huggr-agents Python runtime API, or when deciding between a Python-defined agent and a generated Python wheel from a Rust-defined agent.
Define and run huglets in TypeScript or JavaScript through the huggr-agents WASM runtime, with Node or browser storage, callable tools, streaming events, feedback, and cross-language trace verification. Use when building a Node agent, a browser-hosted agent, or a custom TypeScript host around huggr-core.
Build or modify a Chrome Manifest V3 extension that hosts the Huggr WASM brain, maps browser capability schemas to chrome.* implementations, persists settings and traces in IndexedDB, and calls an OpenAI-compatible model. Use for examples/chrome-extension, custom browser-agent UI, capability dispatch, extension packaging, or WASM/Chrome debugging.
Inspect, replay, verify, compare, and analyze Huggr traces, lineage, costs, tool activity, feedback, and offline insights. Use when an ask fails, a trace no longer verifies, behavior changes after a reducer or policy edit, costs regress, a resumed session looks wrong, or an agent needs trace-driven improvement suggestions.
