跳到主要内容
知仓学习社ZHICANG

context-pack

Generate a compact first-pass repository briefing with context-pack. Use when work starts in an unfamiliar repo, when the user asks for orientation …

不碰外部(只输出文字)无严重或高危命中hashgraph-online/awesome-codex-plugins

它会碰到什么

扫了多少2 个文本文件,3 KB
它会碰到什么不碰外部(只输出文字)
命中总数0 处
命中统计严重 0 · 高 0 · 中 0 · 低 0

这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。

技能内容

Context Pack

Use this skill to turn a repository into a small, prioritized briefing before deeper exploration.

If the context-pack MCP server is installed through this plugin, prefer the MCP tools over shelling out manually:

  • get_context
  • get_changed_context
  • get_file_excerpt
  • init_memory
  • refresh_memory

Workflow

  1. Start with context-pack --cwd <repo> or the get_context MCP tool before a manual tree walk unless the task is already extremely narrow.
  2. Read the generated briefing first. Prioritize guidance docs, manifests, entrypoints, active work, and caveats before opening large source files.
  3. Pick flags based on the task:
  • Active work: context-pack --cwd <repo> --changed-only --no-tree
  • Machine-readable output: context-pack --cwd <repo> --format json
  • Tight prompt budget: context-pack --cwd <repo> --changed-only --no-tree --max-bytes 2000
  • Noisy repo tuning: add --include, --exclude, --max-files, or --max-depth
  1. If the user wants persistent learned notes, use:
  • context-pack --cwd <repo> --init-memory
  • context-pack --cwd <repo> --refresh-memory
  • or the matching init_memory / refresh_memory MCP tools
  1. Only after the briefing is in hand should you move to manual file reads, targeted search, or code changes.

Output Focus

  • Name the repo shape and likely entrypoints.
  • Call out the most useful files to read next.
  • Summarize active work separately from static repo structure.
  • Mention caveats such as missing AGENTS.md, missing README, disabled git, or truncated output.
  • Quote the exact context-pack command you used when it affects the result.

Patterns

  • Orientation: use the default markdown output and summarize what to read first.
  • Reviewing ongoing work: use --changed-only to keep the bundle focused on active files and diffs.
  • Automation: prefer --format json when another tool or script needs the result.
  • Legacy repos: initialize or refresh .context-pack/memory.md when repo-authored docs are weak.

Guardrails

  • Do not substitute a raw tree, blind rg, or random large-file reads for the first pass when context-pack is available.
  • If the generated bundle is too broad or too thin, rerun with tighter budgets or explicit include/exclude globs instead of guessing.
  • Treat repo-type summaries and rankings as heuristics; if the code disagrees, say so explicitly.

想直接用这个技能?

本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。