Github asset manager

GitHub Asset Manager

Manage and improve a GitHub presence. Reads data via gh CLI or GITHUB_TOKEN, runs analysis locally, returns structured markdown reports through stdout so you can discuss results with the user before acting.

What This Skill Does

Can do: read public + (with proper scopes) private GitHub data (profile, repos, stars, activity); generate structured reports on starred and own repos; draft repo metadata recommendations; beautify an existing README while preserving sections; generate multilingual README skeletons + delegate body translation to sub-agents; generate a GitHub Profile README; draft and apply Star Lists classifications.

Cannot do (without explicit user approval): modify any GitHub repo, metadata, Star List, or profile README; assume default languages, tech stacks, featured projects, or classification strategies; push, commit, delete, or overwrite anything automatically.

Default behavior: all commands are read-only. Any write operation requires explicit confirmation.

When to Use

Activate when the user says things like: 整理 GitHub Stars / 分类我的 star; 审计仓库 / 哪些该归档; 帮我生成 GitHub 主页; 润色 / 美化 README; 生成多语言 README / 翻译; 补全仓库信息 / 写 About 描述; 分析 star 的项目.

Standard Workflow

For every command: (1) Authenticate — gh auth status; verify scopes. (2) Clarify intent — what the user wants + which repo/user. (3) Run — node scripts/github-asset-manager.js .... (4) Summarize output — tables/lists, not raw dumps. (5) Propose next steps — concrete. (6) Get explicit approval before any write. (7) Apply manually or via gh api, show the result.

Default Entry Point

For a general GitHub-related request ("帮我看看 GitHub", "整理一下我的 GitHub"), start with audit. If the user's intent is already specific ("分类我的 GitHub Stars", "生成 GitHub 主页"), jump to the matching command.

Commands

Command	Purpose
audit	Run both stars and repos (recommended default)
stars	Analyze starred repositories
repos	Audit own repositories
profile	Generate GitHub Profile README
draft	Completion draft for one repo
beautify	Polish a repo's README
i18n	Generate multilingual READMEs
classify	Generate or apply Star Lists classification

Full syntax, flags, output shapes, per-command workflows: references/commands.md.

Authentication

gh auth status   # preferred
printenv GITHUB_TOKEN   # fallback

If neither is set, ask the user to authenticate. Public data reads with no scope; private needs read:user; classify --apply needs user. Full scope matrix: references/auth-and-scopes.md.

Conversational Patterns

Be a collaborator not a wizard:

• Propose, do not assume. "I detected the README is in Chinese. I recommend Chinese as primary plus English and Japanese. Does that work?"
• Use tables and lists so the user can scan and approve.
• Pause at every decision point — language, tech stack, featured projects, classification strategy, writes.
• Confirm with the exact action. "I will replace README.md and add docs/README.en.md and docs/README.ja.md. Approve?"
• Surface trade-offs. "GitHub only allows one About description. Chinese, English, or combined?"
• Show the result. Commit URLs, file paths, or short summary.

Full patterns + output examples: references/agent-patterns.md.

Error Handling, Caching, Privacy

Failures (401, 404, 403, insufficient scope, empty output), cache lifecycle, push safety, commit conventions, privacy guarantees: references/errors-and-safety.md.

Critical: do not ask the user to paste GITHUB_TOKEN into chat. If a token change is needed, ask them to set it in their environment and restart the session.

References

• references/commands.md — per-command syntax, flags, workflows
• references/auth-and-scopes.md — auth + scope matrix + pre-run checklist
• references/agent-patterns.md — conversational patterns + output examples
• references/errors-and-safety.md — failures, caching, push safety, privacy