القائمة
Initialize this computer with the Kata Engineering framework for Codex and/or Claude.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
| name | kata-init |
| description | Initialize this computer with the Kata Engineering framework for Codex and/or Claude. |
Initialize this computer with the Kata Engineering framework for Codex and/or Claude.
Invoke from the cloned kata-engineering repository:
/kata-init
If the user names a specific agent, configure only that agent. Otherwise, configure every supported agent that appears to be installed on the machine.
| Agent | Global skills directory | Global instruction file (legacy cleanup only) |
|---|---|---|
| Codex | ~/.codex/skills | ~/.codex/AGENTS.md |
| Claude | ~/.claude/skills | ~/.claude/CLAUDE.md |
Setup is symlinks only — global instruction files are never patched. Skills
route themselves through their descriptions and point to the central rules/.
Every skill directory under skills/ is symlinked into each agent's global
skills directory under its own name.
Do not copy Kata skills or rules into individual projects. Global symlinks make every project read the latest version from this repository.
Resolve the Kata Engineering repository root dynamically. Use the current
working directory when it is the repository root; otherwise locate the nearest
parent directory containing both skills/ and rules/.
In shell examples, use:
KATA_ENGINEERING_HOME="$(pwd -P)"
Do not hardcode user-specific absolute paths in reusable repo instructions.
Configure:
~/.codex exists~/.claude existsIf neither agent directory exists and the user did not specify an agent, ask which agent to configure.
For each selected agent, create its global skills directory if needed.
For each selected agent (<dir> is ~/.codex or ~/.claude):
mkdir -p <dir>/skills
for s in "$KATA_ENGINEERING_HOME"/skills/*/; do
ln -sfn "${s%/}" <dir>/skills/"$(basename "$s")"
done
Every skill in the repository is linked — new skills added to the repository
are picked up on the next kata-init run. Verify with file, not ls:
macOS Finder aliases look like files and are not followed by agents.
If a destination already exists as a real file or directory (not a symlink), move it to a timestamped backup under the agent's backup directory before creating the symlink:
~/.codex/backups/~/.claude/backups/Do not delete existing user content.
Earlier versions of Kata Engineering patched global instruction files with an
AGENTS-patch.md routing section. That file no longer exists — skills route
themselves. If present, remove these leftovers:
~/.claude/CLAUDE.md: any @.../kata-engineering/AGENTS-patch.md include line~/.codex/AGENTS.md: any block between <!-- BEGIN KATA ENGINEERING ... -->
and <!-- END KATA ENGINEERING --> markers, and any bare
@.../kata-engineering/AGENTS-patch.md lineRemove only the Kata leftovers. Preserve all other content in those files.
Print a summary:
AGENTS.md and CLAUDE.md files should stay focused on
project-specific constraints.rules/ from the Kata Engineering repository root (two
levels above their real, symlink-resolved location), so projects do not
need local copies of rules/.Implement features, fix bugs, start new projects, or bootstrap project structure. Use for any coding task — writing or modifying code on any project, even small edits.
Review code changes, check project health, verify contract compliance, or refresh living docs (PLAN.md, DECISIONS.md). Use for any review or project health task.
Sweep the project's filenames and folder layout for product-readability. Form independent assumptions first, then propose or apply renames that survive a fresh-eyes test.
Bootstrap a project knowledge base — folder skeleton, entity-page design, SCHEMA/CLAUDE/README scaffolding, an ingest-sources skill, .gitignore, git init, and initial commit. Use when the user wants to set up a knowledge base or queryable source of truth for a stakeholder-heavy project with meetings, emails, and documents.
Process new emails, meeting transcripts, Word documents, presentations, and PDFs dropped into sources/ and distill them into entity pages (people, organizations, workstreams, commitments, decisions, change-requests, risks, assumptions, dependencies, stakeholders, milestones, integrations, budget, meetings). Also updates SOURCES-CHECKLIST.md. Use when the user says "ingest sources", "process new sources", "update the knowledge base", or drops new files into any sources/ subfolder and asks to process them.
Turn a rough project idea into a structured starting point before execution begins. Use this skill whenever starting something new from a vague idea — it interrogates the concept, challenges weak assumptions, and produces a Project Brief, Research Brief, phased project shape, execution plan, and a refined kickoff prompt ready to hand to an agent.