agentsclimarketplace

Ai team kit

Skill siqiyaoyao/ai-team-kit/skill/ai-team-kit

A reusable Agent Skill for keeping Claude Code and Codex aligned through durable project files, handoffs, and workflow memory.

Install
npx -y skills add siqiyaoyao/ai-team-kit --skill ai-team-kit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 16 days oldThe repository was created 16 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Set up or maintain a project folder for an AI coding team by safely creating or merging AGENTS.md, CLAUDE.md, HANDOFF.md, docs/agent_workflow_rules.md, optional .gitignore entries, durable document-update routing rules, and workflow-memory guardrails that prevent agents from re-inventing previously agreed procedures. Use when the user asks to bootstrap a new or existing project for Claude/Codex collaboration, create agent collaboration files, stop pasting the same startup prompt, install an AI Team Kit, prepare a repo for multi-agent coding, reduce agent workflow/prompt drift, preserve project-specific procedures across sessions, or define when agents should update AGENTS.md, HANDOFF.md, README, docs, workflow rules, or skill templates during ongoing work. Defaults to English output unless the user specifies another language.

SKILL.md

4.5 KB, 877 tokens by cl100k_base, as published. Nobody here has run it

AI Team Kit

Use this skill to turn a folder into a project that Claude, Codex, and future coding agents can share without touching application code.

Defaults

  • Target folder: current working directory unless the user provides a path.
  • File language: English by default. If the user specifies Chinese or another language, write all generated collaboration files in that language.
  • Keep AGENTS.md, CLAUDE.md, HANDOFF.md, and docs/agent_workflow_rules.md in one language; do not mix languages within these files.
  • Core file names stay fixed: AGENTS.md, CLAUDE.md, HANDOFF.md, docs/agent_workflow_rules.md.
  • Privacy default: never create a git remote, push, publish, or make a GitHub repo without explicit user approval. If publishing is later requested, default to a private repository.

Workflow

  1. Read references/workflow.md for the full procedure.
  2. Inspect the target folder:
    • Determine whether it is empty, an existing code project, or a knowledge/documentation vault.
    • For existing projects, read real project files such as README, package manifests, config files, folder structure, and entry points.
    • Mark uncertainty as (to confirm) in English, （待确认） in Chinese, or the equivalent phrase in the chosen language.
  3. Protect the current state with local git:
    • If there is no git repo, create .gitignore, run git init, then snapshot the current folder with a baseline commit.
    • If git reports dubious ownership, stop and tell the user to run git config --global --add safe.directory <target folder>; do not bypass it.
    • If git identity is missing (and no global identity exists), set a neutral local placeholder so the baseline can commit, then tell the user to replace it with their own name/email. Never assume a specific person/tool name; never change global git config.
  4. Create or merge the collaboration files:
    • Use assets/AGENTS.en.template.md, assets/CLAUDE.en.template.md, assets/HANDOFF.en.template.md, and assets/AGENT_WORKFLOW_RULES.en.template.md for English.
    • Use the .zh.template.md assets for Chinese.
    • For other languages, adapt the English templates faithfully into the requested language.
    • If files already exist, merge new material into them; do not overwrite user-authored content.
  5. Verify:
    • AGENTS.md stays concise, preferably under 150 lines.
    • CLAUDE.md imports @AGENTS.md and contains only Claude-specific rules.
    • HANDOFF.md contains recent history and leaves current task / next steps blank unless the user supplied them.
    • docs/agent_workflow_rules.md exists or is merged, and AGENTS.md tells agents to check it before nontrivial work.
    • AGENTS.md includes document-update routing so agents know when to update HANDOFF.md, AGENTS.md, CLAUDE.md, README.md, docs/, implementation files, or skills/templates.
    • HANDOFF.md is treated as volatile session state, not the only place for durable project knowledge.
    • No application code changed.
  6. Commit:
    • Stage only AGENTS.md, CLAUDE.md, HANDOFF.md, docs/agent_workflow_rules.md, and .gitignore if this workflow created or updated it.
    • Commit with add agent collaboration files.
    • Do not commit unrelated user changes.
  7. Finish with a one-line summary listing created/merged files and any (to confirm) items.

Resources

  • Full checklist: references/workflow.md
  • English templates: assets/AGENTS.en.template.md, assets/CLAUDE.en.template.md, assets/HANDOFF.en.template.md, assets/AGENT_WORKFLOW_RULES.en.template.md
  • Chinese templates: assets/AGENTS.zh.template.md, assets/CLAUDE.zh.template.md, assets/HANDOFF.zh.template.md, assets/AGENT_WORKFLOW_RULES.zh.template.md

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.