---
name: brewpage-publish
description: Host HTML pages and static sites, publish AI artifacts, and share Markdown, documents, reports or any file with BrewPage. Use when asked to publish, upload, host or share content and return a hosted URL.
license: MIT
---

# BrewPage publishing

Read `upstream/SKILL.md` completely before acting. It contains the existing publishing workflow, security constraints, consent choices, archive review, and exact command arguments. Its seven-file implementation is preserved unchanged from the MIT-licensed upstream skill.

Apply these Qwen Code translations while following that workflow:

- Resolve `${CLAUDE_SKILL_DIR}` to the absolute `upstream` directory beside this file. Resolve `$ARGUMENTS` to the user's actual request. Never expect these Claude variables to exist in the shell.
- Use Qwen Code's native read, glob, shell, and question capabilities for the upstream names `Read`, `Glob`, `Bash`, and `AskUserQuestion`. A normal chat question is valid if no structured question tool exists. Missing tools or answers block the dependent action.
- Claude `model`, tool allowlists, and pre-approval claims are not Qwen settings. Follow Qwen's permission system and never bypass its approval prompts. Preserve the upstream single-command argument escaping and length limits.
- Namespace and password choices require explicit answers before upload; preserve archive confirmation, owner-token secrecy and exact returned URLs.
- For EVERY upstream command, replace `bash ${CLAUDE_SKILL_DIR}/scripts/brewpage-run.sh` with `python3 '/absolute/native/skill/scripts/brewpage-native.py'`, resolving that adapter beside this file. Keep all following arguments unchanged. Never invoke the upstream helper directly.
- The adapter resolves relative input and password-file paths before using its own private working directory. History is isolated at `~/.qwen/brewpage-publish/.claude/brewpage-history.md` (mode 600); report that exact native path. Never read, copy or write a project's or another runtime's Claude history or authentication.
- Input, probe and password paths with `.claude` or `.claude-*` components are forbidden case-insensitively, including aliases resolving there. The helper receives only minimal environment variables, private temporary storage and isolated Git settings.
- The accompanying MCP server is available for separate explicitly requested MCP work; it does not replace or weaken this skill's consent and secure history workflow.

Do not load upstream Claude metadata as a second skill or copy authentication from another client.